Guidance C/2026/6013 is a communication from the Commission, dated 9 November 2026 and published in the Official Journal on 17 November 2026. It explains selected key provisions and is addressed to economic operators as well as to market surveillance authorities, notifying authorities and notified bodies, in order to promote harmonised enforcement. Only the Court of Justice of the EU can give an authoritative interpretation of the CRA, and the examples do not replace a case-by-case assessment. Our recommendation: where you depart from the Commission’s line, you should document the reasons.
The Commission’s CRA Guidance: what the interpretation means for software, devices and open source
On 17 November 2026 the European Commission published its guidance on applying the Cyber Resilience Act in the Official Journal: 257 paragraphs and 67 examples on scope, open source, substantial modification, the support period, important products, remote data processing and reporting obligations. This page puts the clarifications into context — with a decision tree, a case navigator and an integrated gap analysis.
Last updated: November 2026 · Valeri Milke, ISO 27001 & ISO 42001 Lead Auditor
You receive the download link immediately on the page and by email.
C/2026/6013 · Official Journal of 17 Nov 2026 · not binding
- 01Scope
- 02Open source
- 03Risk
- 04Classes
- 05Modification
- 06Support
- 07Remote processing
- 08Reporting
01 Scope · Web app or product? · Locally executed software is a product02 Open source · Who is responsible? · Maintainers and monetisation decide03 Risk · Residual risk acceptable? · Risk tolerance and costs do not count04 Classes · Important or critical? · Exactly one core functionality decides05 Modification · Update substantial? · New or increased risks are the yardstick06 Support · How long? · Expected use time, as a rule at least five years07 Remote processing · Backend part of the product? · Link to a function plus development under the manufacturer’s responsibility08 Reporting · Since when? · Since 11 Sep 2026, early warning within 24 hours
257paragraphs in 9 chapters67examples5use cases
257paragraphs in nine chapters — the consultation draft of March 2026 had 232
67examples, 14 of them new compared with the draft — none replaces a case-by-case assessment
5use cases on remote data processing — from the mobile banking app to the cellular network
24 hdeadline for the early warning to the CSIRT and ENISA after becoming aware of an actively exploited vulnerability or a severe incident — mandatory since 11 September 2026
The Cyber Resilience Act has been in force since 10 December 2024; its reporting obligations have applied since 11 September 2026, and the other obligations of economic operators apply from 11 December 2027. In its guidance, the Commission sets out how it interprets key concepts — from the question of when software is a product, through open source and substantial modification, to core functionality and remote data processing. The guidance is not binding, but it is expressly also addressed to market surveillance authorities and notified bodies. Throughout this page, we distinguish between what the Regulation requires, how the Commission interprets it and what VamiSec recommends.
From the Regulation in 2024 to the guidance in 2026 and the date of application in 2027
The milestones of the Cyber Resilience Act and its guidance — what already applies today and what is still to come.
20 November 2024
CRA in the Official Journal
Regulation (EU) 2024/2847 is published in the Official Journal of the EU. It lays down horizontal cybersecurity requirements for products with digital elements — hardware and software, including their remote data processing.
10 December 2024
Entry into force
The Cyber Resilience Act enters into force. Its obligations apply in stages: notification of conformity assessment bodies from June 2026, reporting obligations from September 2026, all other obligations from December 2027.
3 December 2025
Commission FAQs
The Commission publishes frequently asked questions intended to help economic operators prepare for implementation. The guidance later refers to them expressly.
3 March 2026
Draft out for consultation
The Commission puts the draft of its guidance out to public consultation, which runs until 13 April 2026. The draft comprises 232 paragraphs and 55 examples.
11 June 2026
Notification of conformity assessment bodies
The provisions on the notification of conformity assessment bodies apply. These bodies are needed for conformity assessments with third-party involvement — for important products of class II, for critical products and for class I where no harmonised standard has been applied in full.
11 September 2026
Reporting obligations apply
Manufacturers notify actively exploited vulnerabilities and severe incidents to the coordinating CSIRT and to ENISA — under the Regulation, via the single reporting platform. This also applies to products that were placed on the market before that date.
17 November 2026
Guidance in the Official Journal
Guidance C/2026/6013, dated 9 November 2026, is published in the Official Journal: 257 paragraphs, 67 examples and 5 use cases on remote data processing. It is not binding but reflects the Commission’s interpretation.
11 December 2027
Date of application
All obligations apply: essential cybersecurity requirements, vulnerability handling, conformity assessment, EU declaration of conformity and CE marking. Products placed on the market earlier are covered — apart from the reporting obligations — only in the event of a substantial modification.
11 June 2028
End of old certificates for CRA purposes
Until then, manufacturers can rely, for the cybersecurity risks covered, on EU type-examination certificates and approval decisions under other legislation, for example under the Radio Equipment Directive or the Machinery Regulation, unless they expire earlier. They do not replace a comprehensive CRA risk assessment.
Nine clarifications in the guidance — from its status to the reporting obligations
Click a card to read the clarification — the order follows a product’s path: from the status of the guidance through scope, placing on the market and open source to modifications, core functionality, remote data processing and reporting obligations.
Software that runs on the user’s side — downloaded, installed or supplied in another way — is, in the Commission’s view, a product with digital elements, expressly also as a browser extension or as an app built with web technologies for local execution. Software that runs remotely and that users merely access is not a product in itself: for example, a web application or progressive web app used exclusively in the browser, or a website. It is covered only where, as remote data processing, it supports the function of a product.
Standalone software is considered placed on the market with the first offer of a finished version — for all copies of that version at once, regardless of when they are bought or downloaded. Updates without a substantial modification do not shift this date. Builds for other operating systems or bundles with a different feature set, by contrast, are separate products. Source code can also be a product: anyone who licenses it to customers in the course of a commercial activity places it on the market. Unfinished code for testing or review and demo code in tutorials are not considered placed on the market.
A device and the software from the same manufacturer that it needs in order to be operated, configured or controlled form one product — even if the driver or app comes separately from the app store; both are considered placed on the market at the same time. What matters for scope is not electronics but the exchange of digitally encoded data: signals that merely switch or supply power are not a data connection. Complex systems with long life cycles are not exempt; technical constraints must be documented and mitigated by compensatory measures.
Anyone placing units of a previously developed type on the market from 11 December 2027 does not necessarily have to redesign it. What matters is an up-to-date cybersecurity risk assessment: if it demonstrates appropriate and effective measures, the manufacturer may rely on it. Historical design and test documentation does not have to be recreated, and evidence can be grouped for product families with the same design and risk profile. Conformity assessment, the EU declaration of conformity, CE marking and vulnerability handling nevertheless remain mandatory.
For free and open-source software, the guidance asks two questions: who bears responsibility, and is the software monetised? Responsibility lies with the maintainers who publish and steer releases — not with contributors who merely contribute code. A price, a paid or enterprise version, the monetisation of other services or personal data as a condition of use lead to placing on the market; a donation link, optional support or sponsorship on their own do not. Legal persons that provide sustained support for non-monetised software intended for commercial activities can be open-source software stewards.
A software update is a substantial modification if it introduces new or increased risks that the risk assessment does not cover, or changes the intended purpose — what matters is the risk, not the scope or complexity. Security updates are, as a rule, not substantial, even when they involve major technical changes. The support period reflects the expected use time and is, as a rule, at least five years — a minimum, not a standard. A substantial modification requires a reassessment but does not automatically restart the period.
Whether a product is important (class I or II) or critical depends on exactly one core functionality, assessed objectively and not on the basis of marketing. An integrated operating system or another important component does not automatically make the end product important; modules offered separately are classified individually. Remote data processing is part of the product if two conditions are met together: without it, a product function would be lost, and the manufacturer has developed the software itself or under its responsibility. Who operates it is irrelevant; SaaS bought in from a third party is treated like a component.
Since 11 September 2026, manufacturers have been notifying actively exploited vulnerabilities and severe incidents simultaneously to the coordinating CSIRT and to ENISA: early warning within 24 hours, notification within 72 hours of becoming aware. This also applies to existing products and after the end of the support period, but not retroactively. These very clarifications were missing from the consultation draft of March 2026; also new are the section on web applications, more precise rules on spare parts and provisions on the support period after substantial modifications.
Interactive
Does the CRA affect me — and what does the guidance say?
Up to six questions along the criteria the guidance uses to delimit the scope: software or online service, data connection, open source, placing on the market and core functionality. Check each offering separately — every edition, every separately distributed module. The result is an orientation by VamiSec based on the guidance, not legal advice and not a case-by-case assessment.
Your path
- 1?
1: Do you supply hardware or software that runs on the user’s side — as a device, part, installed app, download, embedded firmware or on a physical medium?
1
Do you supply hardware or software that runs on the user’s side — as a device, part, installed app, download, embedded firmware or on a physical medium?
Yes for devices with or without embedded software, parts such as circuit boards or chips, apps, desktop programs, browser extensions and applications built with web technologies for local execution — also for source code that you supply to customers as a product. No if users only access your software remotely, for example as a web application in the browser, as SaaS or as a website.
Your product is likely to fall under the CRA — with self-assessment under module A
According to the criteria in the guidance, there is much to suggest that you place a product with digital elements on the market and bear manufacturer obligations. Without a core functionality from the lists of important or critical products, self-assessment under the internal control procedure (module A) is sufficient. The reporting obligations for actively exploited vulnerabilities and severe incidents have already applied since 11 September 2026, all other obligations apply from 11 December 2027.
- Set up the cybersecurity risk assessment: according to the guidance, internal risk tolerance, business strategy or costs do not count, and residual risks cannot be shifted to users by means of a warning notice.
- Define product boundaries: the companion app, drivers and your own backend are part of the product if it cannot perform a function without them; manage variants and builds as separate products.
- Derive the support period from the expected use time — as a rule at least five years — and state the end date at the time of purchase, at least with month and year.
- Check the reporting process immediately: early warning within 24 hours and notification within 72 hours to the coordinating CSIRT and to ENISA, also for products that are already on the market.
- Importers and distributors have their own obligations; anyone who substantially modifies a product and makes it available becomes a manufacturer themselves.
Guidance 2.1–2.5 · 7.1Guidance 6.1 · 9.1CRA Art. 13, 14, 32
Important or critical product: a stricter conformity assessment procedure
If the core functionality corresponds to a category on the lists of important or critical products, a stricter procedure applies according to the guidance. Apart from important products placed on the market as free and open-source software, class II and critical products always require third-party involvement — via module B+C, module H or a certification scheme. For class I, the internal control procedure is sufficient only if a relevant harmonised standard whose reference has been published in the Official Journal has been applied in full and covers all risks of the core functionality. The reporting obligations have applied since 11 September 2026, all other obligations apply from 11 December 2027.
- Determine the core functionality objectively for each product and state it unambiguously in the technical documentation — contradictions between advertising, instructions and documentation are a warning sign.
- For class I, monitor the standards situation: self-assessment requires a harmonised standard applied in full whose reference has been published in the Official Journal, or a common specification; as long as these do not exist, we recommend planning for third-party involvement.
- Even with a standard, the risk assessment for the entire product remains mandatory; the presumption of conformity applies only to the risks that the standard covers.
- Important products placed on the market as free and open-source software may use the procedures of the default category.
- Clarify the route to the conformity assessment body early; the provisions on the notification of such bodies have applied since 11 June 2026.
Guidance 6.1–6.3Examples 57–65CRA Art. 7, 8, 32
Your online service is part of the product as remote data processing
On its own, your service is not a product. However, because a product could not perform one of its functions without it and the software was developed by the manufacturer or under its responsibility, it counts, according to the guidance, as remote data processing forming part of the product — with risk assessment, essential cybersecurity requirements, technical documentation and reporting obligations. The manufacturer of the product is responsible, even if a third party operates the service.
- Draw the scope narrowly: covered are the software modules responsible for the function and their interfaces — not server hardware, not HR, CRM or CI/CD systems and not downstream systems without direct interaction with the product.
- Assess downstream systems and third-party cloud platforms as external dependencies and address them with product-level measures, such as authentication and integrity protection.
- State and describe the remote data processing in the technical documentation of each product concerned; the documentation may be reused.
- For a third party’s IaaS or PaaS, obtain evidence from the provider — for example on NIS 2 obligations, under ISO/IEC 27001 or via service level agreements.
Guidance 8.1–8.2Use cases 8.3.1–8.3.5CRA Art. 3(2)
Your service is not in itself a product within the meaning of the CRA
Software that is executed remotely and that users merely access — for example a web application or progressive web app used exclusively in the browser, or an informational website — is not, in the Commission’s view, a product with digital elements in itself. For cloud computing services, the guidance refers to the framework of the NIS 2 Directive. However, do check the environment of your service.
- Check accompanying offerings individually: mobile apps, desktop clients and browser extensions are products; their backend functions may form part of them as remote data processing.
- If you supply your service to manufacturers, they treat it like a third-party component — expect requests for evidence, for example on ISO/IEC 27001, NIS 2 obligations or service level agreements.
- If the offering changes — for example through your own app or a device that needs your service — reassess the scope.
Guidance 2.2 · Examples 5–6Guidance 8.1.3 · 8.2Use case 8.3.3
Without a data connection, your product is outside the scope
The CRA is not linked to electronics but to the ability to exchange digitally encoded information. In the Commission’s view, there is no data connection where electrical or electronic signals only trigger or power a function — for example an output that merely switches on and off without these states representing data or being read by a digital input.
- Document the delimitation: which interfaces and signals exist and why no digitally encoded data flows.
- Take reasonably foreseeable use into account — we recommend including service and diagnostic interfaces.
- Re-check with every new product generation with a radio, bus or app connection.
Guidance 2.5CRA Art. 2(1)
You are unlikely to have CRA obligations for this software
According to the guidance, free and open-source software that is not monetised is not considered placed on the market. Anyone who publishes as an individual or merely contributes code, without controlling releases, roadmap or governance, is not subject to the Regulation to that extent — commit access alone does not change this. Legal persons that do not provide sustained support for software intended for commercial activities can remain entirely free of obligations under the Regulation for this software.
- A donation link is harmless, even if the donations exceed the costs — this is different if releases, binaries or security updates are in effect accessible only to donors.
- Manufacturers that integrate your software bear the due diligence obligation for it themselves; maintainers of components not placed on the market have no obligations towards them.
- Re-check as soon as you monetise — for example with a paid version, with bundled support beyond cost recovery or with data as a condition of use.
Guidance 3.1–3.2 · 3.4Examples 13, 20, 27, 34
You are likely to be an open-source software steward — with a lighter regime of obligations
According to the guidance, a legal person that systematically supports, on a sustained basis, free and open-source software that is intended for commercial activities but not placed on the market is an open-source software steward. Only the specific steward obligations apply to it, not manufacturer obligations. How far the reporting obligations extend depends on the type of your support.
- Only non-technical support such as branding, governance, events or collecting donations: no reporting obligations, but pass on information about actively exploited vulnerabilities to the maintainers in line with your cybersecurity policy.
- IT infrastructure such as repositories, version control or signing keys: notify severe incidents affecting this infrastructure to the CSIRT and ENISA.
- Engineering resources such as developers, review, release management or patches: notify actively exploited vulnerabilities as soon as you become aware of the exploitation — mandatory since 11 September 2026.
- Check the status per project and communicate changes clearly; anyone who monetises later becomes a manufacturer from that point on, not retroactively.
Guidance 3.3–3.3.1 · 9.1Examples 24–26, 28–29, 31–33CRA Art. 3(14), Art. 24
Not yet placed on the market — manufacturer obligations apply once you offer the product
The Regulation is linked to making available in the course of a commercial activity. According to the guidance, unfinished code shared for testing or review, and sample and demo code in tutorials or training material, are not considered placed on the market. As soon as you offer a finished version in the course of a commercial activity — even free of charge — all copies of that version of standalone software are considered placed on the market.
- Under the Regulation, unfinished software such as alpha and beta versions or release candidates may be made available only for the time necessary for testing and gathering feedback.
- Our recommendation: set up risk assessment, documentation and the vulnerability process already during development — the reporting obligations apply from the first placing on the market, all other obligations from 11 December 2027.
- Re-check the status if the offering changes — for example through monetisation, a commercial distribution channel or regular deliveries to customers.
Guidance 2.1 · 2.3Examples 1–2 · 7
- Do you supply hardware or software that runs on the user’s side — as a device, part, installed app, download, embedded firmware or on a physical medium?
- Does a product with digital elements need this online service to perform one of its functions — and did the manufacturer of that product develop the software for it itself or under its responsibility?
- Is it part of the intended purpose or reasonably foreseeable use that the product exchanges digitally encoded data with a device or network — directly or indirectly, via cable, radio or an interface?
- Is the product free and open-source software — with a licence that permits use, modification and redistribution, and with openly shared source code?
- Is the software under your responsibility — you publish it and steer development and releases — and do you monetise this edition?
- Do you, as a legal person, publish this software intended for commercial activities and systematically support its development on a sustained basis?
- Do you offer the finished product on the EU market in the course of a commercial activity — whether in return for payment or free of charge?
- Does the core functionality of your product correspond to a category on the lists of important products (class I or II) or critical products?
Topic navigator
Ten topics in the guidance — from scope to reporting
The guidance explains selected key provisions of the Cyber Resilience Act. The topic navigator arranges its chapters along the product journey — from the question of whether a product is in scope to the support period — and adds three cross-cutting topics. For each topic, you will see typical misconceptions, references, next steps, a review question and evidence. The guidance reflects the Commission’s non-binding interpretation; the structure, weighting and measures are VamiSec’s own assessments.
Product journey
Cross-cutting
Product journey · Guidance 2.2, 2.4–2.6 · Examples 3–6, 8–11
Scope & product boundaries
The guidance draws the line where software reaches the user: anything that is downloaded, installed and executed on the user’s device — app, desktop program, browser extension — is a product with digital elements. A web application that is merely accessed in the browser is not, on that basis alone; it is covered only to the extent that, as remote data processing, it supports a product function. Drivers and companion apps belong to the device if it cannot function in accordance with its intended purpose without them, whatever channel they come through. A genuine data connection is also required; signals that merely trigger or power a function are not enough. Complex systems with long lifecycles remain in scope.
ReferenceGuidance 2.2 to 2.6 (paras 17–32) · Examples 3–6, 8–11 · fn. 10 · CRA Art. 2(1), Art. 3(1), (2) and (4)
Typical misconceptions
- “Web technology is never relevant under the CRA.” — Web-based desktop apps, browser extensions and locally installed clients are products with digital elements. (high)
- “Our web app falls entirely under the CRA.” — An application used exclusively in the browser is not a product in itself; it is covered only to the extent that, as remote data processing, it supports a product function. (medium)
- “The app in the store is a separate product.” — If the device needs it to be operated, configured or used, it forms one product with the hardware and is deemed placed on the market at the same time as the devices. (high)
- “Every device with electronics is in scope.” — What matters is the exchange of digitally encoded information; signals that only trigger or power a function are not a data connection. (medium)
- “Long-lived installations are exempt.” — Complex systems remain in scope; constraints must be documented, and risks mitigated by other means and periodically reassessed. (high)
- “Our product website has to become CRA-compliant.” — A purely informational site is not a product; a portal that issues credentials or tokens required for operation is a different matter. (low)
References
Guidance 2.2 · paras 17–21 · Examples 3–6Guidance 2.4 · para. 25 · Examples 8–9Guidance 2.5 · paras 26–28Guidance 2.6 · paras 29–32 · Examples 10–11 · fn. 10Guidance 8.1.2 · para. 194 (websites)CRA Art. 2(1), Art. 3(1), (2) and (4)
What to do now
- Build a product inventory: record for each offering whether software executes on the user’s side (app, desktop, browser extension, client) or is merely accessed in the browser.
- Assign companion software: treat drivers, apps and configuration tools without which a device cannot function in accordance with its intended purpose as part of the product — regardless of the distribution channel.
- Check the data connection: is digitally encoded information exchanged between sender and receiver, or is a function merely triggered? Document the result with reasons.
- Map web applications and portals against the products: if the remote processing supports a product function, it belongs in the risk assessment and documentation as remote data processing.
- For complex systems, record technical constraints such as legacy protocols or interoperability requirements, define compensatory measures and disclose them in the documentation and user information.
- Where technically feasible, implement the secure protocol and enable it by default; allow older protocols only where they are demonstrably needed for interoperability.
- Periodically reassess constraints during the support period and update the product as soon as they can be lifted or reduced.
Review question for product management
Can you show, with reasons, for every offering whether it is a product with digital elements — and which software, companion apps and remote processing belong to it?
Evidence you can present
- Product and variant inventory with a scope decision for each offering
- Assignment of drivers, apps and companion software to the respective device
- Reasoned assessment of data connection and remote processing
- Documented constraints of complex systems with compensatory measures
- Protocol configuration with secure defaults and justified exceptions
Product journey · Guidance 2.1, 2.3, 2.7 · Examples 1, 2, 7, 12
Placing on the market & legacy designs
Standalone software is placed on the market as soon as a finished version is first offered in the course of a commercial activity — for all copies at once, even if customers buy later. Updates that do not constitute a substantial modification do not shift this date; builds and editions with different components, configurations or functionalities are separate products. Source code supplied to customers as a product is also placed on the market — FOSS code in public repositories generally is not, and neither are unfinished code and demo code. Legacy designs do not need a redesign if the risk assessment demonstrates effective measures; conformity assessment and CE marking remain.
ReferenceGuidance 2.1, 2.3, 2.7 (paras 10–16, 22–24, 33–39) · Examples 1, 2, 7, 12 · fns 7–9 · CRA Art. 3(21) and (22), Art. 4(3), Art. 13(2)
Typical misconceptions
- “Every download is a new placing on the market.” — All copies of a version are deemed placed on the market with the first offer; later downloads count as making available. (medium)
- “Every update is a new product.” — Only a substantial modification leads to a new placing on the market; other iterations do not require a new conformity assessment. (medium)
- “Builds for different platforms are the same product.” — The guidance treats variants with different components, configurations or enabled functionalities as distinct products. (medium)
- “Source code is never a product.” — Anyone who supplies code to customers as a product places it on the market, whether it is source code or machine code. (high)
- “Legacy designs must be redeveloped by December 2027.” — Not necessarily: if the risk assessment demonstrates effective measures, the existing design holds; conformity assessment, EU declaration of conformity and CE marking remain mandatory. (high)
- “For legacy designs we need all historical test reports.” — Earlier design and test documentation does not have to be recreated; a current, documented risk assessment provides the evidence. (low)
References
Guidance 2.1 · paras 10–16 · Examples 1–2 · fn. 7Guidance 2.3 · paras 22–24 · Example 7 · fns 8–9Guidance 2.7 · paras 33–39 · Example 12CRA Art. 3(21) and (22)CRA Art. 4(3) (unfinished software)Blue Guide 2022, Section 2.3
What to do now
- Keep a release register: record for each version and variant the date of first offer and the classification of each subsequent version as substantial or not substantial.
- Treat builds, editions and bundles with different components or functionalities as separate products and, where justified, document them jointly as a product family.
- Classify code deliveries: treat source code licences to customers as products; label public FOSS repositories, pre-release versions and demo code separately.
- Limit alpha, beta and release candidate phases to the time needed for testing and feedback, and clearly label them as pre-release versions.
- For legacy designs, carry out a current risk assessment before the first placing on the market from 11 December 2027 and explain how the existing design mitigates the risks.
- Plan conformity assessment, EU declaration of conformity and CE marking for legacy designs — regardless of whether a redesign is needed.
- Group test evidence for variants with the same design and the same risk profile through representative evidence for the product family.
Review question for product management
Do you know for every version, every variant and every code delivery when it was or will be placed on the market — and which legacy designs you will continue to ship after 11 December 2027?
Evidence you can present
- Release register with the date of first offer for each version and variant
- Reasoned classification of each subsequent version as substantial or not substantial
- Classification of all code deliveries to customers
- Current risk assessment and technical documentation for each legacy design
- Roadmap for conformity assessment, EU declaration of conformity and CE marking
Product journey · Guidance 3.1–3.5 · Examples 13–34 · Figure 1
Free and open-source software
Free and open-source software within the meaning of the CRA requires two things: a licence granting all rights of use, modification and redistribution, and openly shared source code. Responsibility lies with whoever publishes the software and controls releases — not with those who merely contribute code; commit access alone does not change this. Manufacturer obligations arise only with placing on the market: through a price, through monetisation of other services, through mandatory data processing or through donations that in effect buy access. Legal persons that publish FOSS intended for commercial activities and support it on a sustained basis without placing it on the market are stewards with graduated obligations.
ReferenceGuidance 3 (paras 40–89) · Examples 13–34 · Figure 1 · fns 11–15 · CRA Art. 3(14) and (48), Art. 13(5) and (6), Art. 24
Typical misconceptions
- “Open source is exempt from the CRA.” — Only FOSS that is not placed on the market is exempt; monetised editions carry manufacturer obligations, and legal persons can be stewards. (high)
- “A FOSS licence is enough.” — In addition, the source code must be openly shared; code available only to paying customers is not free and open-source software within the meaning of the CRA. (medium)
- “Anyone who submits pull requests to us becomes subject to the CRA.” — Contributors without control over releases, roadmap and governance are not subject to the CRA. (low)
- “The donate button makes us commercial.” — Voluntary donations are harmless, even if they exceed the costs; it is different if releases, updates or security fixes are available only to donors. (medium)
- “Open core is out of scope because the community version is free.” — The paid edition is a separate product placed on the market, with manufacturer obligations. (high)
- “Whoever integrates FOSS is responsible for its compliance.” — Integrators are not responsible for the component itself, but must exercise due diligence, report vulnerabilities to the maintainers and share their own fixes. (medium)
References
Guidance 3 · paras 40–46 (definition, commercial activity)Guidance 3.1 · paras 47–49 · Example 13Guidance 3.2 · paras 50–68 · Examples 14–26Guidance 3.3 · paras 69–84 (stewards)Guidance 3.4–3.5 · paras 85–89 · Examples 27–34 · Figure 1CRA Art. 3(14) and (48), Art. 24
What to do now
- Check each project and each edition for a price, monetisation of other services, mandatory data processing and donation-dependent access — and document the result.
- Keep community and paid versions separate: manufacturer obligations for the monetised edition, a steward check for a legal person’s free version.
- Record roles: who publishes and controls releases, and who only contributes? Commit access alone is not the yardstick.
- Cross-check donation models: do not in effect tie releases, binaries, updates and security fixes to donations if no placing on the market is intended.
- As a steward, determine the type of support — non-technical, infrastructure or engineering — and derive the reporting obligations from it.
- Communicate changes in status clearly: publish the end of sustained support or the start of monetisation with a date.
- For integrated FOSS components, embed due diligence, upstream vulnerability reporting and the sharing of your own fixes in the development process.
Review question for product management
Can you show for every open-source project you publish and every edition whether you are manufacturer, steward or without CRA obligations — and why?
Evidence you can present
- Project list with a role decision for each project and edition
- Documentation of monetisation and donation models
- Governance records on maintainers, releases and contributions
- Description of the steward support and the resulting reporting channels
- Due diligence evidence for integrated FOSS components
Product journey · Guidance 7.1–7.4 · Examples 66–67
Risk assessment & components
The cybersecurity risk assessment is not organisational risk management: the yardstick for residual risks is an appropriate level of cybersecurity of the product, not internal risk tolerance, commercial strategy or cost. Risks can also be treated at product level — through a smaller attack surface, safeguards or a narrower intended purpose — but they cannot be shifted onto users by means of a warning notice. Sourced components are subject to a separate due diligence obligation: determine what the component must deliver and verify this in a risk-based manner. Variants of a product family with the same architecture and the same risks may rely on a joint assessment and documentation.
ReferenceGuidance 7.1–7.4 (paras 156–177) · Examples 66–67 · CRA Art. 13(2), (3) and (5), Annex I Part I
Typical misconceptions
- “We accept residual risks according to our risk appetite.” — Internal risk tolerance, strategy and cost do not count; the yardstick is an appropriate level of cybersecurity of the product. (high)
- “A warning in the manual is enough.” — Risks and responsibility cannot be passed on to users or third parties; user information only supports secure deployment. (high)
- “Too expensive to fix, so it stays as it is.” — Cost alone does not justify untreated risks; what is needed then are changes to design, functionality or intended purpose. (high)
- “The supplier alone is responsible for components.” — The due diligence obligation remains with the manufacturer: determine what the component must deliver and verify in a risk-based manner that it does so. (medium)
- “We have to control the entire back-end environment.” — What is required is to identify external risks and mitigate them on the product, for example through authenticated remote commands and a secure state in the event of outages. (medium)
- “Every variant needs its own risk assessment.” — Not where architecture, security-relevant design, intended purpose and risks are the same; relevant differences such as interfaces or update mechanisms must be reflected. (low)
References
Guidance 7.1 · paras 156–163 · Examples 66–67Guidance 7.2 · paras 164–166Guidance 7.3 · paras 167–173Guidance 7.4 · paras 174–177CRA Art. 13(2), (3) and (5)CRA Annex I Part I
What to do now
- Base the risk assessment for each product on intended purpose, reasonably foreseeable use, operational environment, assets to be protected and expected use time.
- Align acceptance criteria for residual risks with the appropriate level of cybersecurity — remove cost and strategy arguments from the justification.
- Examine treatment options at product level: reduce the attack surface, add safeguards, restrict functionality or define the intended purpose more precisely.
- Write user information on residual risks and limits of use in a manner appropriate to the intended users — as a supplement to measures, not a substitute for them.
- Define security requirements for each sourced component, for example for cryptography, update mechanism and secure communication, and evaluate supplier evidence and your own tests in a risk-based manner.
- Prefer mature platform functions such as the operating system’s encryption and key management where they cover the risks.
- Form product families: record security-relevant differences for each variant and update the joint assessment with every new variant.
Review question for product management
Is every residual-risk decision justified by the product’s level of cybersecurity — without reference to budget, strategy or warning notices — and is it defined for every component what it must deliver?
Evidence you can present
- Risk assessment for each product or product family with operating context and expected use time
- Residual-risk justifications with reference to the essential requirements
- Requirements catalogue and verification evidence for each integrated component
- User information on residual risks and limits of use
- Variant matrix with security-relevant differences
Product journey · Guidance 6.1–6.3 · Examples 57–65 · Figures 2–8
Important & critical products
Whether a product is important (class I or II), critical or part of the default category is determined by its core functionality — exactly one per product, determined objectively by main features and technical capabilities, not by marketing. Ancillary functionalities and integrated components do not change the classification; modules offered separately are each classified in their own right. Class I can rely on module A only if a harmonised standard is applied in full and covers all risks of the core functionality. Class II and critical products always require third-party involvement, except for important products placed on the market as FOSS. The presumption of conformity extends only as far as the standard.
ReferenceGuidance 6 (paras 136–155) · Examples 57–65 · Figures 2–8 · fns 21–27 · CRA Art. 7, 8, 27, 32 and 64(3) · Implementing Regulation (EU) 2025/2392
Typical misconceptions
- “Our product has several core functionalities.” — For the choice of procedure, every product has exactly one; it must be clearly identified in the technical documentation. (medium)
- “An integrated firewall makes our device a class II product.” — What counts is the core functionality of the whole; the mere integration of an important or critical component does not change the classification. (medium)
- “With the right product description we end up in the default category.” — The assessment is objective; misrepresentation by overemphasising or downplaying functionalities is not permitted and can trigger fines. (high)
- “Class I always needs a third-party assessment.” — Not if a harmonised standard is applied in full and covers all risks of the core functionality; module A is then possible. (medium)
- “With a harmonised standard, the presumption covers the whole product.” — It applies only to the risks covered by the standard; additional functionalities need their own evidence. (high)
- “A suite is classified by its strongest module.” — Modules offered separately are products in their own right with their own classification; only modules supplied solely as integrated parts count towards the overall product. (medium)
References
Guidance 6 · paras 136–138 · fns 21–23Guidance 6.1 · paras 139–146 · Examples 57–61Guidance 6.2 · paras 147–152 · Examples 62–63 · fns 24–25Guidance 6.3 · paras 153–155 · Examples 64–65 · fns 26–27CRA Art. 7, 8, 27 and 32Implementing Regulation (EU) 2025/2392 (technical description of the categories)
What to do now
- Determine the core functionality of each product and each separately offered module — based on the instructions for use, promotional material and technical documentation.
- Compare the core functionality with the technical descriptions in Implementing Regulation (EU) 2025/2392 and record the reasoning, even if the result is negative.
- Check product texts, instructions for use and technical documentation for inconsistencies with the defined core functionality.
- Define the conformity assessment procedure: module A for the default category; for class I only with a harmonised standard applied in full that covers the core functionality, otherwise with third-party involvement.
- For class II and critical products, clarify early on which conformity assessment body is available and when.
- Prepare a standards gap analysis: which risks does the standard cover, and which additional functionalities need their own measures and evidence?
- Once references of harmonised standards are published in the Official Journal, re-examine coverage and the choice of procedure.
Review question for product management
Is exactly one core functionality documented with reasons for every product — and can the chosen conformity assessment procedure be derived from it without gaps?
Evidence you can present
- Core functionality analysis for each product and separately offered module
- Comparison with the categories of Implementing Regulation (EU) 2025/2392
- Reasoned choice of procedure: module A, module B+C, module H or certification scheme
- Standards gap analysis with additional measures for functionalities not covered
- Consistency check of promotional material, instructions for use and documentation
Product journey · Guidance 4.1–4.4 · Examples 35–51
Substantial modification & spare parts
A substantial modification exists where a change after placing on the market affects compliance with the essential cybersecurity requirements or changes the intended purpose that was assessed. For software updates, what counts is not their scale but whether they introduce new or increased risks that the risk assessment does not cover. Security updates are generally not substantial modifications, unless they fundamentally change purpose, data flows or external dependencies. Repairs are generally not substantial modifications either; identical spare parts supplied specifically for repair are exempt. Anyone who substantially modifies someone else’s product and makes it available is considered its manufacturer.
ReferenceGuidance 4 (paras 90–124) · Examples 35–51 · fns 16 and 20 · CRA Art. 2(6), Art. 3(30), Art. 21, 22 and 69(2)
Typical misconceptions
- “Every feature update is a substantial modification.” — Not if the functionality was foreseen and assessed in the risk assessment or does not alter the risk profile. (medium)
- “Small changes are uncritical.” — The yardstick is the potential deterioration of the risk profile, not the scale; even a “remember me” feature can be a substantial modification. (high)
- “Security updates are never substantial modifications.” — Generally not, even where they involve major technical changes; they are, however, if they fundamentally change purpose, data flows or external dependencies. (high)
- “All spare parts are exempt.” — Only identical parts manufactured according to the same specifications and supplied specifically for repair; the security-relevant characteristics are decisive. (medium)
- “Rebuilding purchased modules only modifies someone else’s product.” — Anyone who integrates components into a product of their own places a new product on the market, with full manufacturer obligations. (high)
- “A substantial modification requires a complete reassessment.” — The focus is on the modified parts; existing documentation and tests may be reused. (low)
References
Guidance 4 · paras 90–92Guidance 4.1–4.2 · paras 93–102 · Examples 35–39Guidance 4.3 · paras 103–113 · Examples 40–50Guidance 4.4 · paras 114–124 · Example 51 · fn. 16CRA Art. 2(6), Art. 3(30)CRA Art. 21, 22 and 69(2)
What to do now
- Make update classification a mandatory step in the release process: new threat vectors, new attack scenarios, changed likelihood, changed impact.
- Anticipate planned functionalities in the risk assessment, assess them and back them with mitigation measures — they then generally do not trigger a substantial modification when activated.
- Check security updates separately: do they change the intended purpose, data flows, external interfaces or third-party dependencies?
- Set up a spare parts process: identity assessment based on security-relevant characteristics, and evidence of the supply context such as the product reference and the after-sales channel.
- Check your own modifications to third-party products and your integration projects for a change of role to manufacturer.
- In the event of a substantial modification, focus the conformity assessment on the modified parts and justify the reuse of existing evidence.
- Update the risk assessment and technical documentation after every release — regardless of the outcome of the classification.
Review question for product management
Can you show in a traceable way for every release whether it was a substantial modification — and what consequences you drew from that?
Evidence you can present
- Update classification for each release with a check against the criteria
- Updated risk assessment and technical documentation
- Identity assessments and supply-context evidence for spare parts
- Role analysis for modifications to third-party products and for integrations
- Conformity assessment of the modified parts for substantially modified versions
Product journey · Guidance 5 and 5.1 · Examples 52–56
Support period
The support period reflects the expected use time of the product — measured against reasonable user expectations, the nature of the product and relevant Union law. Five years is a minimum, not a default; the period may be shorter only if the expected use time is shorter. The end date must be indicated at the time of purchase, at least the month and year. Software manufacturers may remediate vulnerabilities in the latest version only if the upgrade is available free of charge and without additional costs. A substantial modification requires a reassessment but does not automatically restart the period.
ReferenceGuidance 5 and 5.1 (paras 125–135) · Examples 52–56 · fns 17–20 · CRA Art. 13(8), (10) and (19)
Typical misconceptions
- “Five years of support is the standard.” — Five years is the minimum; if the product is used for longer, the period must be set correspondingly longer. (high)
- “We choose three years because it is more economical.” — Shorter than five years only if the expected use time is genuinely shorter. (high)
- “Every substantial modification starts five new years.” — No: if the factors determining the use time stay the same, the remaining period applies, even if it is less than five years. (medium)
- “With the latest version, every obligation for older versions ends.” — Only remediation may be limited to the latest version; coordinated disclosure, information sharing and reporting obligations remain. (medium)
- “The end of support can go in the small print.” — The end date must be indicated clearly and understandably at the time of purchase, at least the month and year. (medium)
- “Any upgrade effort counts as additional costs.” — Testing, configuration and routine work do not count; new hardware or fundamental infrastructure changes do. (low)
References
Guidance 5 · paras 125–131 · Examples 52–53 · fns 17–19Guidance 5.1 · paras 132–135 · Examples 54–56 · fn. 20CRA Art. 13(8) (determination, minimum duration)CRA Art. 13(10) (latest version)CRA Art. 13(19) (end date, notification)
What to do now
- Derive the expected use time for each product: user expectations, nature of the product, hardware durability, support periods of comparable products and of sourced core components.
- Determine and justify the support period — use five years only as a minimum, and shorter periods only with evidence of shorter use.
- Indicate the end date, at least month and year, at all points of sale and in the product information; prepare the end-of-support notification where technically feasible.
- Review the upgrade policy: can users of earlier versions obtain the latest one free of charge and without additional costs? Only then limit remediation to it.
- After every substantial modification, check whether the factors determining the use time have changed, and confirm or recalculate the period.
- Align the lifecycles of sourced components such as the operating system, libraries and chips with your own support period.
Review question for product management
Can you explain for every product why the stated support period corresponds to the expected use time — and where customers find the end date at the time of purchase?
Evidence you can present
- Derivation of the expected use time for each product
- Determined support period with the end date in the purchase information
- Upgrade and version policy with a cost assessment for users
- Documented reassessments after substantial modifications
- Alignment with the lifecycles of sourced components
Cross-cutting · Guidance 8.1–8.3 · Use cases 8.3.1–8.3.5 · Figures 9–10
Remote data processing
Software that processes data remotely is also part of the product — but only if two conditions are met at the same time: without the processing, the product could not perform one of its functions, and the software was designed and developed by the manufacturer or under its responsibility. Who operates it is irrelevant; the manufacturer’s own servers and private clouds can also qualify. Third-party SaaS is not the manufacturer’s remote data processing, but it is treated like a component. Telemetry used purely for statistics, internal IT, CI/CD and test systems are not included. Downstream back-end systems remain external dependencies.
ReferenceGuidance 8 (paras 178–208) · Use cases 8.3.1–8.3.5 · Figures 9–10 · fns 28–29 · CRA Art. 3(1) and (2)
Typical misconceptions
- “Any use of the cloud is remote data processing.” — Only if a product function depends on it and the software was developed by the manufacturer or on its behalf to its specifications. (high)
- “Our back-end is none of the CRA’s business.” — Your own back-end software without which a product function would fail is part of the product — even on third-party IaaS or PaaS. (high)
- “The operator decides.” — What counts is design and development; if a third party operates the manufacturer’s own solution, the manufacturer remains responsible. (medium)
- “Then third-party SaaS is none of our business.” — It is not the manufacturer’s remote data processing, but it must be included like a third-party component in the risk assessment, product-level measures and due diligence. (medium)
- “Our entire corporate IT becomes subject to the CRA.” — Human resources, CRM, CI/CD pipelines as well as test and red-teaming systems are not remote data processing solutions. (low)
- “Telemetry is always part of the product.” — Not if it serves only statistics or future product development; its risks nevertheless belong in the risk assessment. (low)
References
Guidance 8 · paras 178–183 · fns 28–29Guidance 8.1.1–8.1.3 · paras 184–202 · Figure 9Guidance 8.2 · paras 203–208Guidance 8.3.1–8.3.5 (use cases) · Figure 10Guidance 7.3 · para. 169 (back-end risks)CRA Art. 3(1) and (2)
What to do now
- Inventory all remote processing for each product and classify it using both key questions: would a function fail without it? Was the software developed by the manufacturer or under its responsibility?
- Draw the boundary: modules interacting directly with the product and their interfaces belong to the remote data processing; downstream systems are external dependencies.
- Describe the remote data processing and the IaaS or PaaS services used in the technical documentation of each product concerned, including details of the contracted service.
- Treat third-party SaaS like components: assess integration risks, implement product-level measures such as isolation and content validation, and obtain evidence from the providers.
- Use reusable evidence from providers, for example on NIS 2 obligations, DORA, European certifications or ISO/IEC 27001 and 27017.
- Embed security guarantees and information on changes in service level agreements, and review the risk assessment when providers make changes.
- Include remote data processing in vulnerability handling and reporting processes.
Review question for product management
Is it documented for every product which remote processing belongs to the product, which is treated as a third-party component and where the boundary with the rest of the back-end lies?
Evidence you can present
- Inventory of remote processing classified against both key questions
- Architecture diagram showing the boundary and external dependencies
- Technical documentation describing remote data processing and IaaS or PaaS dependencies
- Due diligence evidence and service level agreements of third-party providers
- Risk assessment including downstream systems and the product environment
Cross-cutting · Guidance 9.1–9.2 · applicable since 11 September 2026
Reporting obligations & vulnerabilities
Since 11 September 2026, manufacturers have had to notify actively exploited vulnerabilities and severe incidents simultaneously to the CSIRT designated as coordinator and to ENISA — including for products already on the market and after the end of the support period. The deadlines run from becoming aware, that is, from a reasonable degree of certainty after an immediate initial assessment: early warning within 24 hours at the latest, notification within 72 hours, complete report 14 days after a corrective or mitigating measure is available (vulnerability) or one month after the 72-hour notification (incident). Exploitation already known before the cut-off date is not reported retroactively.
ReferenceGuidance 9.1 (paras 209–221) and 9.2.1–9.2.3 (paras 222–240) · fn. 30 · CRA Art. 13(6), Art. 14, Art. 24(3), Annex I Part II
Typical misconceptions
- “The reporting obligations only apply from December 2027.” — They have applied since 11 September 2026, including for products already sold and after the end of support. (high)
- “The clock starts with the first indication.” — The deadline starts on becoming aware with a reasonable degree of certainty; the initial assessment must, however, be carried out immediately. (high)
- “Every vulnerability in a library must be reported.” — Only what is actively exploited in your own product is notified; otherwise, report upstream to the maintainers and handle it as usual. (medium)
- “A vulnerability is only known once it is in a database.” — Non-public information, your own testing, including AI-powered testing, and prominent reports in reliable media also make it known. (medium)
- “Nothing may be shipped with a known exploitable vulnerability.” — The manufacturer decides on a risk basis according to severity, exploitability and impact; the risks of a postponement also count. (medium)
- “Regular testing means the same pentest once a year.” — What is required is a regular review of whether new threats or vulnerabilities call for new tests — proportionate to risk rather than in a rigid cycle. (low)
References
Guidance 9.1 · paras 209–221 · fn. 30Guidance 9.2.1 · paras 222–229Guidance 9.2.2 · paras 230–237Guidance 9.2.3 · paras 238–240CRA Art. 13(6), Art. 14, Art. 24(3)CRA Annex I Part II
What to do now
- Set up a reporting process with roles, deputies and on-call cover that ensures the early warning, notification and complete report are submitted on time and plans for informing impacted users.
- Prepare reporting channels: the Regulation provides for ENISA’s single reporting platform for this purpose; clarify responsibilities, access and report contents in advance.
- Bundle intake channels for reports from customers, researchers, authorities and the media, carry out an initial assessment of every report immediately and document the point in time of becoming aware.
- Include products already on the market and products past the end of their support period in the reporting scope.
- For third-party components, analyse reachability and exploitation in your own product, report vulnerabilities upstream through the maintainers’ channels and share your own fixes.
- Before every placing on the market, take a documented, risk-based release decision on known exploitable vulnerabilities.
- Maintain a test plan proportionate to risk and adapt it to new threats or vulnerabilities, including regression testing after changes.
Review question for product management
Could you notify an actively exploited vulnerability in one of your products within 24 hours today — with a documented point in time of becoming aware and coordinated user information?
Evidence you can present
- Reporting process with deadlines, roles and on-call plan
- Log of initial assessments with the point in time of becoming aware
- Register of upstream reports and shared fixes
- Release decisions on known exploitable vulnerabilities
- Risk-based test plan with change history
Cross-cutting · Guidance 9.3 · outlook in 1.2
Interplay with other legislation
Vehicles of categories M, N, O and L that are subject to the type-approval rules, and components exclusively designed and constructed for integration into them, do not fall under the CRA. Generic components or parts sold to anyone through open channels do — regardless of statements on intended use. Existing EU type-examination certificates, for example under the Radio Equipment Directive or the Machinery Regulation, hold only for the risks they cover and until 11 June 2028 at the latest; they do not replace a comprehensive CRA risk assessment. The Commission is considering further guidance, for example on the interplay with the AI Act and DORA.
ReferenceGuidance 9.3.1 (paras 241–246) and 9.3.2 (paras 247–257) · fn. 31 · Guidance 1.2 (paras 6, 9) · CRA Art. 2(2)(c), Art. 69(1)
Typical misconceptions
- “Automotive supplier parts are exempt across the board.” — Only components exclusively designed and constructed for such vehicles; generic parts and open distribution channels bring a product within the CRA. (high)
- “The note ‘for motor vehicles only’ is enough.” — What counts are the objective conditions under which the product is made available, for example whether anyone can order through general retail outlets or online channels. (medium)
- “Our RED certificate proves CRA compliance.” — It holds only for the risks covered; vulnerability handling, data minimisation or attack surface may remain open. (high)
- “For the CRA, the certificate is valid until its own expiry.” — For CRA purposes only until 11 June 2028 at the latest, even if it remains valid for longer under other legislation. (medium)
- “With a certificate, there is no need for a CRA risk assessment.” — A comprehensive risk assessment remains mandatory; gaps must be assessed and mitigated. (high)
- “The guidance settles everything for AI products and financial entities.” — It concerns only the CRA; the Commission is considering further guidance on the interplay with the AI Act and DORA. (low)
References
Guidance 9.3.1 · paras 241–246Guidance 9.3.2 · paras 247–257 · fn. 31Guidance 1.2 · paras 6 and 9CRA Art. 2(2)(c), Art. 69(1)Regulation (EU) 2019/2144, Regulation (EU) No 168/2013, Delegated Regulation (EU) 2025/1535Delegated Regulation (EU) 2022/30 (radio equipment), Regulation (EU) 2023/1230 (machinery)
What to do now
- Check automotive components: exclusively designed and constructed for type-approved vehicles? Restrict distribution to the automotive supply chain and document this.
- Plan the CRA route for generic or openly sold components — regardless of statements on intended use.
- Inventory existing type-examination certificates: risks covered, validity and products placed on the market between 11 December 2027 and 11 June 2028.
- Prepare a gap analysis between certificate and CRA, for example on vulnerability handling, data minimisation and attack surface, and close the gaps.
- Plan the transition by 11 June 2028 at the latest: CRA conformity assessment for all products that will continue to be placed on the market after that date.
- Track overlaps with the AI Act, DORA, NIS 2, radio equipment law and machinery law in a single register of obligations, and monitor further Commission guidance.
Review question for product management
Do you know for every product which legal acts apply alongside the CRA, which certificates hold and until when — and where gaps to the CRA remain?
Evidence you can present
- Scope memo for vehicle components with evidence of the distribution channels
- Register of existing type-examination certificates with the risks covered
- Gap analysis between certificate and CRA with an action plan
- Transition plan for products placed on the market up to 11 June 2028
- Register of obligations across all relevant legal acts
Scope & product boundaries
Case navigator
All 67 examples and 5 use cases in the guidance — searchable and filterable
The Commission illustrates its interpretation with 67 numbered examples and five use cases on remote data processing. The navigator summarises each case in our own words: the outcome in the Commission’s view and what it means for manufacturers. Filter by chapter and outcome, search for product types such as app, printer, router or library, and copy the filtered list as a checklist. The examples are not binding and do not replace a case-by-case assessment.
67numbered examples in chapters 2 to 7 of the guidance
5use cases on remote data processing: mobile banking, thermostat, e-Reader, industrial robot, cellular network
22cases on free and open-source software — the largest group
17cases on substantial modifications, repairs, spare parts and software updates
Chapter
Outcome
72 of 72 cases
ExampleCaseOutcome & implicationsSection
Example 1ScopeSoftware version by download, copies sold on different daysPlacing on the marketBoth copies of version 1.0.0 are deemed placed on the market on the day of the first offer — including the one bought two weeks later. For standalone software, what counts is the first offer of a finished version, not the individual sale; later downloads count as making available. For manufacturers, this means: the date of placing on the market is tied to the version — record it for each version, because conformity and the support period are linked to it.Section 2.1In scope
Example 2ScopeFollow-up version 1.0.1 without substantial modification — new date of placing on the market?Placing on the marketNo. The copy of version 1.0.1 is also deemed placed on the market on the day of the first offer of 1.0.0, because the update is not a substantial modification. Iterations that are not substantial modifications do not require a new conformity assessment procedure and do not shift the date. For manufacturers, this means: the classification of each update determines whether there is a new placing on the market — document it for each release.Section 2.1In scopeModification & spare parts
Example 3ScopeMobile app from the app store, installed on the smartphoneSoftware as a product with digital elementsThe app is supplied to the user and executes on the user’s device — it is a product with digital elements. It may fall within the scope of the CRA if it is placed on the market in the course of a commercial activity, whether in return for payment or free of charge. For app providers, this means: free of charge does not automatically mean out of scope; the commercial context is decisive. Also check which remote processing the app needs for its functions.Section 2.2In scope
Example 4ScopeDesktop application built with web technologies, packaged for local installationSoftware as a product with digital elementsAn application built with web technologies is also a product with digital elements as soon as it is supplied for local installation and runs on the user’s device; if it is placed on the market in the course of a commercial activity, it may fall within the scope of the CRA. For manufacturers, this means: the technology is not decisive — what counts is whether the software executes on the user’s side. According to the guidance, the same applies to browser extensions.Section 2.2In scope
Example 5ScopeWeb application used only in the browser, or locally installed client?Software as a product with digital elementsThe web application used only through the browser is not a product and does not fall within the scope of the CRA, unless it supports the functionality of a product. The locally installed client, by contrast, is a product; if it relies for its functions on processing at a distance that qualifies as remote data processing, that processing is part of the product. For SaaS providers, this means: check for each offering whether there is a client — and which back-end it depends on.Section 2.2Out of scopeIn scope
Example 6ScopeWebsite that only displays informationSoftware as a product with digital elementsA website that merely provides information and does not support any product function is not a product with digital elements and does not fall within the scope of the CRA. It is different where a website enables a function — for example as a portal that issues credentials or tokens required for operation; it can then be remote data processing. For manufacturers, this means: product pages and help centres stay out of scope, while login and control portals should be scrutinised.Section 2.2Out of scope
Example 7ScopeSource code licence to a customer that adapts and compiles the codeComputer codeCompany A places the source code, supplied as a text file, on the market and is subject to the CRA, even though company B still has to adapt and compile it. According to the English version, A is not responsible for the CRA compliance of B’s subsequent adaptations; the German version states the opposite at this point. For manufacturers, this means: code licences to customers are products — the boundaries of responsibility should be clearly set out in the contract.Section 2.3In scopeManufacturer role
Example 8ScopeNetwork printer with drivers for download on the manufacturer’s websiteCombination of hardware and software forming a product with digital elementsPrinter and drivers together constitute a single product with digital elements, although they are supplied through different channels: without the drivers, the printer cannot fulfil its intended purpose. Such software is deemed placed on the market at the same time as the devices. For manufacturers, this means: drivers, configuration and management software belong in the device’s risk assessment, technical documentation and vulnerability handling.Section 2.4In scopeManufacturer role
Example 9ScopeFitness wearable with the manufacturer’s separately downloaded smartphone appCombination of hardware and software forming a product with digital elementsWearable and app together constitute a single product, because they are designed and intended to operate together — even if the app comes separately from the app store. It displays measurements and history and configures the device. For manufacturers, this means: the companion app is part of the device’s conformity assessment, documentation and vulnerability handling; app updates are therefore updates of the product.Section 2.4In scopeManufacturer role
Example 10ScopeInteroperability with legacy systems requires an older, less secure protocolComplex systemsThe less secure protocol may be implemented where this is necessary for the interoperability that forms part of the intended purpose and the risks are identified and mitigated through other means. If supporting both is technically feasible, the Commission expects the secure protocol to be enabled by default, and the older one only where interoperability requires it. For manufacturers, this means: document and disclose the justification, compensatory measures and default setting.Section 2.6Obligations & evidence
Example 11ScopeAutomation system with microcontrollers purchased before the date of applicationComplex systemsThe manufacturer includes the purchased microcontrollers in the system’s risk assessment; if it identifies no specific risks, it integrates one of them and ensures that the system as a whole meets the essential requirements. For manufacturers, this means: legacy stocks of components are not an obstacle — what counts is the assessment within the overall product. If risks emerge, compensatory measures and documentation are needed.Section 2.6Manufacturer roleObligations & evidence
Example 12ScopeMicrocontroller developed before 11 December 2027 and still placed on the market afterwardsProducts with digital elements designed before the CRA entered into applicationNo redesign is needed if a current risk assessment shows that the original design incorporates appropriate and effective measures. Historical design and test documentation does not have to be recreated; for variants based on the same design and with the same risk profile, representative evidence for the product family may suffice. For manufacturers, this means: legacy designs need a risk assessment and technical documentation — vulnerability handling, conformity assessment and CE marking remain mandatory.Section 2.7In scopeObligations & evidence
Example 13Open sourcePull or merge request with a security patch for someone else’s FOSS projectDetermining if free and open-source software is under one’s responsibilityWhoever submits the request is a “contributor” and is not subject to the CRA — even as an employee of a company. They have contributed code but control neither development nor releases or distribution; that is done by the maintainers, who review, merge and publish. For manufacturers, this means: your developers’ contributions to other projects do not make you the manufacturer of those projects; commit access alone does not change this.Section 3.1Out of scopeOpen source & stewards
Example 14Open sourceFree FOSS marketplace app, revenue from advertising, commissions or subscriptionsMonetisation of other services or requiring the processing of personal dataThe app is considered placed on the market in the course of a commercial activity, even though it is free of charge and open source: the publisher monetises other products and services through it. Manufacturer obligations therefore apply to the publisher in full. For providers, this means: it is not the price of the software that is decisive, but whether money is made through it — the business model and revenue streams belong in the scope check of every FOSS project.Section 3.2.2In scopeManufacturer role
Example 15Open sourceFree FOSS VPN with paid additional servers or dedicated IP addressesMonetisation of other services or requiring the processing of personal dataThe VPN is considered placed on the market because the publisher monetises other services through the free application — here, additional servers and dedicated IP addresses. For providers, this means: freemium models in which the open-source software opens up access to paid services entail manufacturer obligations for the software. Check premium servers, upgrades and in-app purchases as possible monetisation.Section 3.2.2In scopeManufacturer role
Example 16Open sourceFree FOSS fitness app, usable only with data processing for advertisingMonetisation of other services or requiring the processing of personal dataThe app is considered placed on the market because its use is conditional upon the processing of personal data for advertising or analytics unrelated to security, compatibility or interoperability. For providers, this means: mandatory data processing as a condition for use acts like a price. Data processing that serves exclusively security, compatibility or interoperability does not have this consequence.Section 3.2.2In scopeManufacturer role
Example 17Open sourceFOSS operating system plus a paid version with support and optimisationSupport servicesThe paid version with technical assistance or performance optimisation is considered placed on the market in the course of a commercial activity. Free and paid editions are different products; a legal person’s free version can trigger steward obligations. For providers, this means: enterprise editions carry full manufacturer obligations — even if a functionally equivalent version is freely available.Section 3.2.3In scopeManufacturer role
Example 18Open sourceFree FOSS command line tool, consultancy and training offered separately for a feeSupport servicesThe software is not considered placed on the market: it can be downloaded freely, and consultancy, training or installation support are offered separately and optionally for a fee. The decisive factor is whether access to the software itself, including its maintenance, is conditioned on remuneration. For providers, this means: professional services around free software are possible as long as download and updates do not depend on a purchase.Section 3.2.3Out of scope
Example 19Open sourceService provider installs third-party FOSS on a customer’s serverSupport servicesThe service provider does not place the software on the market as long as it does not substantially modify it: it does not publish it and only helps install it on the customer’s on-premises server. For IT service providers and integrators, this means: installation and technical assistance for third-party FOSS do not trigger manufacturer obligations — but a substantial modification of the software would change that.Section 3.2.3Out of scope
Example 20Open sourceFOSS tool with voluntary donations, access to code and updates independent of donationsDonationsNo commercial activity, no placing on the market: the donations are voluntary, and the software, source code and updates are open to everyone. The guidance considers a mere call for donations harmless even if the donations exceed the costs. For projects, this means: calls for donations are unproblematic as long as access, releases and security updates are not tied to a donation and donations do not bring contractual benefits or exclusive advantages that go beyond community perks.Section 3.2.4Out of scope
Example 21Open sourceFOSS licence, but releases and security updates only for donorsDonationsThe donations are de facto a condition for access and therefore amount to a price: anyone who does not donate does not get the current version. The software is considered placed on the market — despite the free and open-source licence. For projects, this means: anyone who puts downloadable releases or security updates behind a donation takes on the manufacturer role with full obligations. Check donation models for de facto access barriers.Section 3.2.4In scopeManufacturer role
Example 22Open sourceSource code public, binaries, updates and security fixes only for donorsDonationsHere too, the software is considered placed on the market: the donations are linked to access to essential aspects of the product and thus constitute remuneration, even though the source code is public. For projects, this means: open source code alone does not protect against manufacturer obligations if pre-compiled builds, regular updates or guaranteed security fixes are available only in return for a donation.Section 3.2.4In scopeManufacturer role
Example 23Open sourceManufacturer funds a new FOSS feature from an individual developerFinancing of free and open-source softwareThe developer does not place the software on the market: the funded feature ends up in the codebase, openly shared and freely available, and the funding is irrelevant to the commercial nature. If the funding manufacturer integrates the FOSS into its product, it must fulfil the due diligence obligation for the component. For manufacturers, this means: paid features in upstream projects make neither the developer nor you the manufacturer of the FOSS.Section 3.2.5Out of scopeObligations & evidence
Example 24Open sourceNot-for-profit browser, monetised via search engine partnershipsNot-for-profit entities set up to achieve not-for-profit objectivesThe browser is not considered placed on the market, even though it is directly monetised: all earnings after costs go to not-for-profit objectives. The publishing legal person is, however, subject to the obligations for open-source software stewards. For not-for-profit organisations, this means: no manufacturer obligations, but steward obligations if they meet the characteristics of a steward — including a cybersecurity policy and, depending on the type of support, reporting obligations.Section 3.2.6Out of scopeOpen source & stewards
Example 25Open sourceLegal person publishes a UI library and uses it in its own productIntegration by other manufacturersThe library is not considered placed on the market: it is not monetised, but it is intended for integration into products. The publishing legal person is its steward — and the manufacturer of its own product placed on the market that contains the library. For companies, this means: roles apply per project; the same organisation can be steward for one FOSS and manufacturer for another product.Section 3.2.7Out of scopeOpen source & stewards
Example 26Open sourceOperating system manufacturer publishes non-monetised reference librariesIntegration by other manufacturersFor the reference libraries, which are intended for integration into other products and are not monetised, the legal person is the steward; for the operating system placed on the market, it remains the manufacturer. For platform providers, this means: classify SDKs and reference libraries project by project, and for steward projects define a cybersecurity policy and reporting channels according to the type of support.Section 3.2.7Open source & stewardsOut of scope
Example 27Open sourceIndividual developer with a donation link, companies integrate and donate voluntarilyIllustrative scenariosThe software is not placed on the market; the individual developer has no CRA obligations, even if companies support them with voluntary donations. The integrating companies must fulfil their due diligence obligation when integrating it. For manufacturers, this means: donations to maintainers do not shift responsibility — checking the component remains your task, as part of your own manufacturer obligations.Section 3.5Out of scopeObligations & evidence
Example 28Open sourceNot-for-profit foundation maintains a FOSS component, companies contribute developer timeIllustrative scenariosBecause the foundation is organised as a not-for-profit entity, the component is not considered placed on the market; the foundation is the steward with the corresponding obligations. Companies that integrate it and voluntarily contribute developer time do not thereby become responsible for the component, but must exercise due diligence when integrating it. For manufacturers, this means: participating in foundation projects does not establish a manufacturer role for the component.Section 3.5Open source & stewardsObligations & evidence
Example 29Open sourceCompany publishes its own component separately, without any monetisationIllustrative scenariosThe component is not considered placed on the market; company A is not its manufacturer but its steward. Companies B, C and D, which integrate it and contribute developer time, must fulfil their due diligence obligation when integrating it. For companies, this means: anyone who publishes and maintains their own building blocks as open source bears steward obligations — the manufacturer obligations remain for their own products that contain them.Section 3.5Open source & stewardsObligations & evidence
Example 30Open sourceCompany offers FOSS as a paid version with support and optimisationIllustrative scenariosCompany A is considered the manufacturer of the software; if it were organised as a not-for-profit entity, it would, according to the guidance, be a steward instead of a manufacturer. Companies B, C and D, whose developers contribute while control remains with A, have no CRA obligations for this software; if they integrate it, the due diligence obligation applies. For providers, this means: paid editions with added value constitute placing on the market — contributions from third parties do not change this.Section 3.5Manufacturer roleObligations & evidence
Example 31Open sourceCompany maintains FOSS for integrators, a partner offers support independentlyIllustrative scenariosCompany A is the steward of the software: it maintains it for the long term, does not charge for it, does not process personal data collected through the product and does not sell support tied to its publication. Company B contributes code and sells support independently of this — without CRA obligations for this software. For service providers, this means: independent support for third-party FOSS does not make you a manufacturer.Section 3.5Open source & stewardsOut of scope
Example 32Open sourceNot-for-profit entity, maintenance funded by research grants, features developed with manufacturersIllustrative scenariosThe not-for-profit entity is the steward of the component. A manufacturer that co-develops and funds features in partner projects does not thereby become the manufacturer of the component; if it integrates the component into its product, the due diligence obligation applies to it. For manufacturers, this means: joint feature projects with FOSS entities are possible without taking on responsibility for the component.Section 3.5Open source & stewardsObligations & evidence
Example 33Open sourceNot-for-profit entity funded by membership fees maintains a FOSS SDKIllustrative scenariosThe not-for-profit entity is the steward of the SDK. The members, who pay fees and provide developers, are not responsible for the SDK’s compliance with the CRA. Manufacturers that use the SDK as a component must fulfil their due diligence obligation. For member companies, this means: membership and participation do not create responsibility for the SDK — due diligence for their own use remains.Section 3.5Open source & stewardsObligations & evidence
Example 34Open sourceIndividual developer publishes a library on a public package repositoryIllustrative scenariosNeither the individual developer nor the package repository has CRA obligations in this scenario; the manufacturer that integrates the library, downloaded free of charge, must fulfil its due diligence obligation. For manufacturers, this means: for libraries from public package sources, the checks are your responsibility — provenance, maintenance status and known vulnerabilities belong in the due diligence. According to the guidance, the scenarios are purely hypothetical.Section 3.5Out of scopeObligations & evidence
Example 35Substantial modificationServer repair: defective RAM replaced with a better-performing onePhysical repairsNot a substantial modification: compliance with the essential requirements is not affected, and the higher performance remains within the use considered in the risk assessment. Better-performing replacement parts do not in themselves trigger a substantial modification. For manufacturers, this means: repairs are generally uncritical, but they should be briefly checked as to whether compliance or the intended purpose is affected.Section 4.1Modification & spare parts
Example 36Substantial modificationIdentical replacement module for connected controllers from 2026 and 2028Spare partsIn both cases the spare parts exemption applies: the identical module, manufactured according to the same specifications, is not itself subject to the CRA, even though it is a product with digital elements — regardless of whether the controller was placed on the market before or after the date of application. The repair is not a substantial modification. For manufacturers, this means: document the identity and the spare-part supply route, and the replacement stays lean.Section 4.2Out of scopeModification & spare parts
Example 37Substantial modificationReplacement chip with different cryptography and new secure boot for a controller from 2026Spare partsNot identical: the different cryptographic implementation and the changed secure boot mechanism affect the cybersecurity properties. The exemption does not apply; the replacement chip is a product in its own right that is subject to the CRA and whose compliance must be assessed in light of its intended purpose — including interoperability with the legacy device. For manufacturers, this means: plan for discontinuations early and document requirements that cannot be met together with compensatory measures.Section 4.2In scopeObligations & evidence
Example 38Substantial modificationNew wireless module with a different chipset and new firmware, same security characteristicsSpare partsThe replacement module can be considered identical: same function, same communication protocols and security mechanisms; the different chipset and updated firmware do not change any characteristics relevant to cybersecurity. It falls under the exemption and not under the CRA. In the consultation draft, the outcome was still the reverse. For manufacturers, this means: identity is measured by security characteristics, not by identical components — assess and document.Section 4.2Out of scope
Example 39Substantial modificationReplacement CPU or complete replacement PLC for an automation systemSpare partsBoth options fall under the spare parts exemption — replacing a component of the PLC as well as replacing the whole PLC, provided it is identical. The supply context is also decisive: the replacement is supplied through the after-sales service channel as a spare part, and the target system is clearly identified. For manufacturers, this means: supply spare parts through service channels with a reference to the product and keep the evidence available for market surveillance.Section 4.2Out of scopeObligations & evidence
Example 40Substantial modificationDashboard for displaying machine data gains control functionsSoftware updates as substantial modificationsSubstantial modification: a situational awareness tool becomes a product that exercises operational control over other devices — adjusting operating parameters, restarting machines after faults. This intended purpose was not envisaged in the risk assessment. For manufacturers, this means: the leap from displaying to controlling is a typical trigger for a new placing on the market, including conformity assessment of the modified parts.Section 4.3Modification & spare parts
Example 41Substantial modificationConsumer app for personal data gains automated profilingSoftware updates as substantial modificationsSubstantial modification: the app shifts from user-controlled information management to an automated decision-making system that prioritises, suppresses or recommends content without user intervention. This was not envisaged in the risk assessment. For manufacturers, this means: assess automation and profiling functions that shift the intended purpose early in the roadmap — otherwise they trigger a new placing on the market.Section 4.3Modification & spare parts
Example 42Substantial modificationMessaging app gains the group chat that was already assessedSoftware updates as substantial modificationsNot a substantial modification: group messaging, together with administrator controls and moderation tools, was foreseen in the original design and assessed in the risk assessment, including the more complex message routing. The update remains within the scope of the original intended purpose. For manufacturers, this means: anyone who assesses planned functionalities in advance can roll them out later without a new conformity assessment.Section 4.3Modification & spare parts
Example 43Substantial modificationDisabled control loops of a monitoring system are activated as assessedSoftware updates as substantial modificationsNot a substantial modification: the risk assessment explicitly covered the future activation of the automated control loops, together with safeguards such as operator override and fail-safe states; the update activates both as assessed. For manufacturers, this means: include functionalities that are already built in but disabled in the risk assessment — with the safeguards that take effect on activation.Section 4.3Modification & spare parts
Example 44Substantial modificationUpdate adds “remember me” with locally stored tokensSoftware updates as substantial modificationsSubstantial modification, even though the feature is small: locally stored authentication tokens bring unassessed risks such as token theft, unauthorised access and session hijacking, and affect compliance with the essential requirements. For manufacturers, this means: what counts is not the scale but the potential deterioration of the risk profile — convenience features also belong in the update classification.Section 4.3Modification & spare parts
Example 45Substantial modificationNew diagnostics feature exports logs, sensitive data unencryptedSoftware updates as substantial modificationsSubstantial modification: the seemingly minor feature collects and stores sensitive operational data in unencrypted form and creates unassessed, unmitigated risks of data exposure with a potentially significant impact on the risk profile. For manufacturers, this means: check logging, diagnostics and export functions for exposure risks before release and plan the protection of the collected data from the outset.Section 4.3Modification & spare parts
Example 46Substantial modificationSecurity update fixes a buffer overflow or an authentication bypassSoftware updates as substantial modificationsNot a substantial modification: the update only changes the internal implementation, does not affect the intended purpose, introduces no new exposure and serves exclusively to reduce risk. For manufacturers, this means: classic vulnerability fixes can be shipped without a new conformity assessment; the risk assessment and technical documentation must nevertheless be kept up to date.Section 4.3Modification & spare parts
Example 47Substantial modificationHardening update: firewall rules, ports, password policies, mandatory MFASoftware updates as substantial modificationsNot a substantial modification: hardening existing security configurations may affect how users configure or access the product, but it does not change the intended purpose and serves solely to improve security. Mandatory MFA here concerns a functionality that was already available or foreseen. For manufacturers, this means: hardening is no reason for restraint — but inform users in good time about changes to access.Section 4.3Modification & spare parts
Example 48Substantial modificationDeprecated algorithm disabled, stronger existing alternative activatedSoftware updates as substantial modificationsNot a substantial modification: the risk assessment covered all cryptographic options and their future deprecation, with cryptographic agility, internal key management and compatibility testing. The update introduces neither new external dependencies nor altered data flows. For manufacturers, this means: cryptographic agility in design and risk assessment pays off — the change of algorithm becomes a routine update.Section 4.3Modification & spare parts
Example 49Substantial modificationSecurity update replaces local encryption with a cloud encryption serviceSoftware updates as substantial modificationsSubstantial modification, although security-driven: instead of encrypting locally, the product uploads all files to a remote encryption service operated by the manufacturer. The intended purpose changes fundamentally and in a manner not foreseen. For manufacturers, this means: classify security updates too — anyone who rebuilds architecture and data flows beyond the risk assessment generally triggers a new placing on the market and is likely to create new remote data processing.Section 4.3Modification & spare parts
Example 50Substantial modificationNew cryptographic protocol requires a third party’s external key management serviceSoftware updates as substantial modificationsSubstantial modification: dependencies and data flows change materially; new external interfaces and a reliance on third-party services not considered in the risk assessment bring new risks. The fact that the update responds to cryptographic weaknesses does not change this. For manufacturers, this means: new third-party dependencies are a warning sign — assess them in advance and examine alternatives without external interfaces.Section 4.3Modification & spare parts
Example 51Substantial modificationAgricultural monitor built from purchased, modified modules with proprietary firmwareSubstantial modifications carried out by a person other than the original manufacturerNot a substantial modification of someone else’s product, but a new product: the company integrates purchased microcontroller modules and connectivity components with proprietary firmware and sensors and places the result on the market under its own name. It is the manufacturer and must comply with the CRA for the product as a whole. For integrators, this means: full obligations — but evidence from the component manufacturers may be used.Section 4.4.1Manufacturer roleIn scope
Example 52Support periodSmartphone with free operating system upgrades during the support periodSupport periodThe manufacturer may limit the addressing and remediation of vulnerabilities to the latest operating system version, as long as earlier versions can be upgraded free of charge and without additional costs. Coordinated vulnerability disclosure and information sharing remain mandatory for the entire support period. For manufacturers, this means: the upgrade strategy must ensure free updates without new hardware; otherwise the relief does not apply.Section 5Obligations & evidence
Example 53Support periodEnterprise software with substantially modified versions every few monthsSupport periodThe manufacturer may discontinue remediating vulnerabilities for earlier versions once users can switch to the latest version free of charge; testing and configuration effort do not count as additional costs, but new hardware or fundamental infrastructure changes do. Each version needs its own declared support period. For manufacturers, this means: document the upgrade path, the cost assessment and the remaining vulnerability obligations.Section 5Obligations & evidence
Example 54Support periodRobot vacuum cleaner gains new cleaning modes via software update (substantial modification)Substantial modifications and the support periodThe support period aligns with the remaining expected use time as originally determined: the update changes neither the physical durability of the hardware nor users’ expectations as to the product’s lifetime. A substantial modification requires a reassessment but does not automatically restart the period. For manufacturers, this means: for pure software updates to hardware-bound products, the end date generally remains unchanged.Section 5.1Modification & spare partsObligations & evidence
Example 55Support periodIndustrial machine: cloud back-end rebuilt, with new APIs and data flowsSubstantial modifications and the support periodHere, the cloud back-end is remote data processing and therefore part of the product; its rearchitecting is a substantial modification. However, the nature of the product and user expectations remain the same, so the support period aligns with the remaining original expected use time. For manufacturers, this means: back-end rebuilds can trigger a new placing on the market without extending the support period.Section 5.1Modification & spare partsObligations & evidence
Example 56Support periodPLC gets a new computing platform with a significantly longer lifetimeSubstantial modifications and the support periodThe manufacturer recalculates the support period: with a new generation of processor, memory and runtime environment, the nature of the product and user expectations change, and the PLC remains in productive use for longer. The use-time criteria now indicate a longer duration. For manufacturers, this means: revitalising hardware modifications or a rewrite of the core software can require longer support.Section 5.1Modification & spare partsObligations & evidence
Example 57Important & critical productsSoftware as an abstraction layer over the hardware — core functionality of an operating system?Core functionalityYes. Initialisation, process and memory management, input and output, scheduling, resource allocation and system interfaces make the software the central runtime environment; the technical documentation and the manufacturer’s statements confirm this. It has the core functionality of an operating system as described for that category in Implementing Regulation (EU) 2025/2392. For manufacturers, this means: the classification follows from features and documentation, not from the product name.Section 6.1Classification
Example 58Important & critical productsSmartphone with integrated operating systemCore functionalityThe smartphone does not have the core functionality of an operating system: as a whole, it serves, for example, communication and access to information and services. The mere integration of an operating system does not change the classification of the end product. For manufacturers, this means: what counts is the product as a whole, not its components. The example leaves open which category the smartphone belongs to instead.Section 6.1Classification
Example 59Important & critical productsSOAR software that can also perform SIEM functionsCore functionalitySOAR software generally does not have the core functionality of a SIEM system: although it can collect, correlate and present data, its core functionality substantially exceeds this, with incident response as a core element. For manufacturers, this means: the mere ability to perform the functions of a category is not enough for classification. Whether another category applies must be checked separately.Section 6.1Classification
Example 60Important & critical productsLog collection and visualisation tools without correlationCore functionalitySuch tools generally do not have the core functionality of a SIEM system: they ingest log data and show basic dashboards, but do not correlate data or provide actionable security insights. They thus fall substantially short of the category. For manufacturers, this means: describe the range of functions realistically — and check whether further development later crosses the threshold to SIEM.Section 6.1Classification
Example 61Important & critical productsSecurity suite with SIEM, intrusion detection and analytics modules, also available individuallyCore functionalityBecause the modules can also be subscribed to individually, each is a standalone product with its own core functionality: the SIEM module is an important product of class I, the intrusion detection system one of class II, and the analytics module belongs to the default category. For providers of modular suites, this means: plan conformity assessment for each separately distributed module; only modules supplied solely as integrated parts count towards the overall product.Section 6.1Classification
Example 62Important & critical productsAntivirus software (class I) with disk-cleaning and anti-tracking functionsConformity assessment for important and critical products with digital elementsThe manufacturer may use the internal control procedure for the entire product: it assesses the risks of the whole product, applies a harmonised standard covering the risks of the antivirus core functionality and treats the additional functions with measures of its own. For manufacturers, this means: for class I, module A is attainable if the standard is applied in full and covers the core functionality — the gap regarding additional functions must be documented.Section 6.2ClassificationObligations & evidence
Example 63Important & critical productsRouter (class I) with integrated firewall componentConformity assessment for important and critical products with digital elementsWhat counts is the router core functionality of the overall product, not the integrated firewall component. With a risk assessment covering everything, a harmonised standard for the router risks and additional measures for the firewall functionality — for instance a harmonised standard for firewalls — the manufacturer may use the internal control procedure for the whole product. For manufacturers, this means: integrated components of higher classes do not shift the classification.Section 6.2ClassificationObligations & evidence
Example 64Important & critical productsAntivirus software: standard covers only the core functionality, not the additional functionsImplications for presumption of conformityThe internal control procedure remains permissible, but the presumption of conformity applies only to the risks of the core functionality — not to the disk-cleaning and anti-tracking functions, whose risks the standard does not cover. For manufacturers, this means: demonstrate and document compliance with the essential requirements for additional functions by other means; the presumption extends only as far as the standard.Section 6.3ClassificationObligations & evidence
Example 65Important & critical productsAs in Example 64, but an updated standard now also covers the disk-cleaning functionImplications for presumption of conformityThe internal control procedure remains permissible; the presumption of conformity now extends to the core functionality and the disk-cleaning function, but not to the anti-tracking function. For manufacturers, this means: keep track of the status of standards on an ongoing basis — an update can extend the reach of the presumption and reduce your own evidence effort for additional functions. For functions not covered, your own evidence remains necessary.Section 6.3ClassificationObligations & evidence
Example 66Risk assessmentIndustrial sensor without protection against physical attacks, only for trusted environmentsOn the evaluation and treatment of cybersecurity risksTo mitigate the risks, the manufacturer limits the intended purpose to trusted environments without unauthorised physical access and informs users clearly, in a manner appropriate to the intended users, about this limitation. The example contains no explicit statement on conformity. For manufacturers, this means: a more precisely defined intended purpose can be part of risk treatment — but a warning notice alone does not shift the risk onto users.Section 7.1Obligations & evidence
Example 67Risk assessmentProfessional app uses the operating system’s encryption and key managementOn the evaluation and treatment of cybersecurity risksAgainst unauthorised access to locally stored sensitive data, the manufacturer relies on the operating system’s native encryption and key management functions instead of a proprietary development — because these are mature and regularly maintained. The guidance presents this as a risk-based measure, without an explicit statement on conformity. For manufacturers, this means: proven platform functions are often the more secure choice.Section 7.1Obligations & evidence
Use case 8.3.1Remote data processingBanking app: banking interface, ledger and a SaaS provider’s support chatUse cases for remote data processing solutions — Mobile banking applicationThe self-developed banking interface is remote data processing: without it, the app can neither authenticate nor submit instructions. The account management and ledger systems are not, because the app does not interact with them directly; they remain external dependencies. A SaaS provider’s support chat is not remote data processing, but it must be treated like a third-party component — with product-level measures and due diligence.Section 8.3.1In scopeOut of scopeObligations & evidence
Use case 8.3.2Remote data processingSmart thermostat with app, own remote processing on a third party’s IaaSUse cases for remote data processing solutions — Smart thermostatRemote data processing: without the remote processing, the smart functions are missing, and the software was developed under the manufacturer’s responsibility — even though it runs on third-party IaaS. The manufacturer documents the remote data processing and the IaaS dependency, including details of the contracted service, and includes both in the risk assessment. For manufacturers, this means: obtain security evidence from the IaaS provider, for example on meeting NIS 2 obligations.Section 8.3.2In scopeObligations & evidence
Use case 8.3.3Remote data processinge-Reader software stores purchased books in a third-party SaaS storage serviceUse cases for remote data processing solutions — e-ReaderNot remote data processing of the manufacturer: although the storage service is necessary for a function, it was developed neither by the manufacturer nor under its responsibility, and it is open to all customers. The manufacturer treats it like a component: assess the risks and implement secure authentication, encryption and integrity protection. For manufacturers, this means: select and integrate SaaS providers with the same diligence as third-party components.Section 8.3.3Out of scopeObligations & evidence
Use case 8.3.4Remote data processingIndustrial robot, camera-based part recognition in its own remote service on third-party IaaSUse cases for remote data processing solutions — Industrial robotThe software of the remote service developed by the manufacturer is remote data processing: without the position calculation from the camera images, the robot cannot pick up parts. It runs on a third party’s IaaS. The manufacturer documents the remote data processing and the IaaS dependency, takes both into account in the risk assessment and implements the requirements on the product, including the remote processing. For manufacturers, this means: select IaaS providers with due care.Section 8.3.4In scopeObligations & evidence
Use case 8.3.5Remote data processingSmartphone uses a telecommunications operator’s 5G networkUse cases for remote data processing solutions — Cellular networkNeither remote data processing nor a third-party component: the network is only a communication channel; whether it is in operation says nothing about whether the product is working correctly. The same applies to ethernet cables, routers or Wi-Fi. Because no software is integrated into the product, there are no due diligence obligations towards the network operator. For manufacturers, this means: network risks belong in the risk assessment, not in supplier vetting.Section 8.3.5Out of scope
Counting and summaries: VamiSec, as of 21 November 2026, based on the Commission guidance on the application of the Cyber Resilience Act (C/2026/6013, Official Journal of the EU of 17 November 2026). The cases are summarised in our own words; the outcome filters are a VamiSec classification. The guidance is not binding; only the Court of Justice of the EU can give an authoritative interpretation, and the examples do not replace a case-by-case assessment. Where the official German version differs from the English one (for instance in Example 7), the summaries follow the English version.
In depth
The CRA guidance in detail: 14 chapters from scope to the roadmap for 11 December 2027
These 14 chapters cover the status and structure of the guidance, placing software on the market, web applications and source code, hardware with companion software, legacy designs, free and open-source software, substantial modification and spare parts, the support period, important and critical products, risk assessment and components, remote data processing, reporting obligations and the interplay with other legal acts — in each case distinguishing between what the Commission interprets, what the Regulation requires and what VamiSec recommends.
What the guidance is — and what it is not: mandate, status, structure
In Communication C/2026/6013, the Commission sets out how it understands key concepts of the Cyber Resilience Act. The guidance is not binding, but it is intended to help economic operators comply and to support harmonised enforcement by the authorities. This chapter sets out the mandate, origins, status and structure of the guidance and shows what has changed compared with the consultation draft.
The “Commission guidance on the application of Regulation (EU) 2024/2847 (Cyber Resilience Act)” is a Communication from the Commission. It was published on 17 November 2026 in the Official Journal of the European Union, C series, under number C/2026/6013; according to the Official Journal metadata, the underlying Commission document C(2026) 7851 is dated 9 November 2026. Its subject is the Regulation on horizontal cybersecurity requirements for products with digital elements. The Regulation entered into force on 10 December 2024, its reporting obligations for actively exploited vulnerabilities and severe incidents have applied since 11 September 2026, and it will apply in full from 11 December 2027.
257consecutively numbered paragraphs in nine sections
67examples, 14 of them new compared with the consultation draft
5use cases on remote data processing, from the mobile banking application to the cellular network
4mandatory topics that the Regulation sets for the guidance
The mandate: guidance required by the Regulation itself
The Regulation requires the Commission to publish guidance to assist economic operators in applying it — with a particular focus on facilitating compliance by microenterprises and small and medium-sized enterprises. The Commission must address at least four topics: the scope, with particular regard to remote data processing solutions and free and open-source software; the notion of support periods; the interplay with other Union law; and the concept of substantial modification. The guidance is not addressed only to manufacturers, importers and distributors. It is also expressly intended to support the activities of market surveillance authorities, notifying authorities and notified bodies, so that the Regulation is enforced in a harmonised manner across the Union.
From the FAQs through the consultation to the final version
As early as 3 December 2025, the Commission had published frequently asked questions (FAQs) intended to help with preparations for implementation; the guidance itself refers to them. The Commission put the draft guidance out to public consultation from 3 March to 13 April 2026, also involving the Expert Group on Cybersecurity of products with digital elements. The final version followed in November 2026. Our assessment of how the two documents relate: the guidance is the more detailed document and the one expressly required by the Regulation; it does not say whether or how the Commission will adapt the FAQs. Anyone who has based scope or role decisions on the FAQs should check them against the guidance.
What the guidance is not
The guidance is not binding for economic operators or other actors. Only the Court of Justice of the European Union can give an authoritative interpretation of the Regulation; in the guidance, the Commission sets out its interpretation in order to support compliance. The examples illustrate this interpretation but do not replace a case-by-case assessment — according to the guidance, such an assessment will always be necessary. Nor does the guidance cover the Regulation in its entire scope. It explains the rationale of certain key provisions and how they could be implemented in practice. And it concerns only the Cyber Resilience Act, not other Union legal acts.
The Commission may consider further guidance — this is not a commitment — for example for manufacturers that are also subject to other Union harmonisation legislation. As examples, it mentions the interplay with the AI Act, Regulation (EU) 2024/1689, and with DORA, Regulation (EU) 2022/2554 on digital operational resilience for the financial sector. What the guidance already says about vehicle legislation and existing EU type-examination certificates is covered in Chapter 13.
Structure: nine sections, 67 examples
| Section of the guidance | What it covers | Examples | Chapter on this page |
|---|---|---|---|
| 1 Introduction | Objectives of the Regulation, mandate, consultation, status of the guidance | — | 1 |
| 2 Scope | Placing on the market, software as a product, computer code, hardware with software, data connection, complex systems, legacy designs | 1–12 | 2 to 5 |
| 3 Free and open-source software | Responsibility, placing on the market and monetisation, open-source software stewards, contributors, scenarios | 13–34 | 6 |
| 4 Substantial modifications and spare parts | Physical repairs, spare parts, software updates, consequences of a substantial modification | 35–51 | 7 |
| 5 Support period | Determination and minimum duration, new versions, relationship with substantial modification | 52–56 | 8 |
| 6 Important and critical products | Core functionality, conformity assessment, presumption of conformity | 57–65 | 9 |
| 7 Cybersecurity risk assessment | Risk treatment, due diligence for components, product families | 66–67 | 10 |
| 8 Remote data processing | Assessment criteria, reliance on third-party solutions, use cases | 5 use cases | 11 |
| 9 Additional elements | Reporting obligations, vulnerability handling, interplay with other legislation | — | 12 and 13 |
Section titles abbreviated. Chapter 14 of this page has no counterpart in the guidance: it is a roadmap and is marked as a VamiSec recommendation.
What has changed compared with the draft
Compared with the consultation draft of March 2026, the final version has grown considerably: from 232 to 257 numbered paragraphs and from 55 to 67 examples. In total, 14 examples are new, one has been dropped and two have been merged; the text is around 23 per cent longer. Entire sections are new: on software as a product with digital elements, on the consequences of a substantial modification, separately for third parties and the original manufacturer, on the relationship between substantial modification and the support period, and on effective and regular tests. The legal nature is unchanged: the final version is not binding either.
| Topic | What the final version changes | Our assessment |
|---|---|---|
| Software as a product | New section: software executed locally is a product, expressly including browser extensions and locally installed applications built with web technologies. Software that executes remotely and is merely accessed is not a product on that basis alone; websites are covered only as remote data processing (Examples 3 to 6). | Web applications used exclusively in the browser are not a product in their own right — they remain covered only to the extent that, as remote data processing, they support the functionality of a product. |
| Free and open-source software | What matters is a price charged for the software itself; the free version is not placed on the market, even in the open-core model; paid support does not in itself constitute placing on the market. Example 13 on contributors and statements on changes in a steward’s status are new. | Manufacturer obligations apply to the monetised edition. For the free version, legal persons are subject to the steward obligations; natural persons are subject to none. |
| Spare parts | The exemption applies only where the part is specifically supplied for repair or to extend durability, as apparent from the context of the supply. Whether a part is “identical” is measured by its functional role and security-relevant characteristics. Example 38 reaches the opposite conclusion to the draft; Example 39 is new. | A part with a different chipset or new firmware can remain exempt if the security-relevant characteristics stay the same and it is demonstrably supplied as a spare part. |
| Software updates | New functionalities that do not, in themselves, alter the risk profile are not a substantial modification either, nor are security updates that introduce no new risks — even where they involve significant technical changes. Materially changed product boundaries or dependencies, such as new data flows or externally reachable interfaces, are treated as warning signs. | Feature updates without new or increased risks do not trigger a new conformity assessment. Architecture, data flows and interfaces become the key checkpoints. |
| Consequences of a substantial modification | Restructured by third parties and original manufacturer. Third parties bear manufacturer obligations only for the modified part, as long as the cybersecurity of the product as a whole is not negatively affected; anyone who assembles components into a product of their own is integrating (Example 51). The same proportionality applies to the original manufacturer for products placed on the market before 11 December 2027. | A substantial modification to an existing product does not automatically require full compliance of the entire product — it does, however, if it negatively affects the product’s cybersecurity as a whole. |
| Support period | New section: a substantial modification requires a reassessment, but neither automatically resets nor automatically extends the period. The period is recalculated where the factors determining the expected use time change (Examples 54 to 56). | Releases with substantial modifications do not extend the period by themselves; a support period must nevertheless be specified for every substantially modified version. |
| Remote data processing | Clarified boundary: what is covered are the software modules responsible for the functionality of the product, together with the interfaces those modules use with external services. Further back-end systems with which the product does not directly interact are not included. | The boundary expressly belongs in the technical documentation. According to the guidance, downstream systems remain external dependencies to be addressed in the risk assessment. |
| Reporting obligations and tests | Clarified: the Regulation’s reporting obligations have applied since 11 September 2026 to all products in scope, including those placed on the market before 11 December 2027 and after the end of the support period; there is no retroactive reporting where the active exploitation was already known before 11 September 2026. A new section covers effective and regular tests without a fixed schedule. | Reporting processes must cover the entire portfolio, including existing products. You should document the point at which you became aware in each case. |
A selection of changes verifiable in the wording. Further clarifications concern, among other things, separately offered modules, risk treatment and known exploitable vulnerabilities. All example numbers refer to the final version; from Example 3 onwards, practically all numbers have shifted compared with the draft.
Placing software on the market: version, copy, variant
When a device is placed on the market is a matter of well-established practice. For standalone software that can be downloaded any number of times, the Commission considers that separate guidance is needed. The guidance takes the version as its reference point: what counts is its first offering, and every variant counts as a separate product.
Making available and placing on the market
The Regulation applies to products with digital elements made available on the Union market. Making available on the market means any supply for distribution or use in the course of a commercial activity, whether in return for payment or free of charge; placing on the market is the first making available. Citing the “Blue Guide”, the Commission’s 2022 guide on the implementation of EU product rules, the guidance emphasises that both concepts refer to each individual product, not to a type of product — whether it was manufactured as an individual unit or in series. For machinery or radio equipment this is well established, and the “Blue Guide” contains its own examples on this.
For standalone software, the notion of the individual unit fits only to a limited extent. There are no physical limits on production or stock, and every download creates a new, identical copy. The Commission resolves this as follows: once the manufacturing phase is complete and the software is offered to prospective users on the Union market, the manufacturer is regarded as having manufactured multiple copies of the same product. The software is placed on the market at the moment of the first offering for distribution or use. Subsequent downloads or distribution of this version are instances of making available — as long as the version is not modified in a way that affects compliance with the Regulation.
All copies of a version: one date
Each copy remains an individual product with digital elements. According to the guidance, however, all copies of a version are considered to be placed on the market at the same time, regardless of when possession or use of the individual copy is transferred to a customer. Two examples in the guidance show the consequences, including for updates.
| Example | Sequence of events | Date of placing on the market |
|---|---|---|
| Example 1 | Company A first supplies version 1.0.0 of its software X for distribution via its website on 1 January 2028. Customer 1 purchases a copy on the same day, customer 2 on 15 January 2028. | Both copies: 1 January 2028 |
| Example 2 | As in Example 1, but on 15 January 2028 version 1.0.1 is released, which does not constitute a substantial modification. Customer 2 purchases a copy of version 1.0.1 on 30 January 2028. | Copies of 1.0.0 and 1.0.1: 1 January 2028 |
The dates are example dates from the guidance.
Example 2 illustrates the second rule. Subsequent iterations are considered newly placed on the market only if they constitute a substantial modification. Iterations without a substantial modification do not require a new conformity assessment procedure and leave the date of placing on the market unchanged. The guidance addresses when an update is substantial in a separate section; Chapter 7 summarises the criteria.
Variants are separate products
Variants are a different matter. Where software differs in its included components, configurations or enabled functionalities — the guidance mentions builds for different operating systems and bundles with differing feature sets — these are not copies of the same product. They should be treated as distinct products with digital elements for the purposes of placing on the market, each with its own date of placing on the market. This need not mean additional documentation work: according to the guidance, variants of a product family with the same architecture, the same security-relevant design, the same intended purpose and the same risks can share one risk assessment, one set of technical documentation and one conformity assessment procedure (Chapter 10).
Copy
Every further download or sale of a version already offered. Not a new placing on the market but making available; the date of the first offering remains (Example 1).
New version without substantial modification
Bug fixes and updates that are not a substantial modification. No new conformity assessment procedure; the date of placing on the market remains unchanged (Example 2).
New version with substantial modification
Considered newly placed on the market. Before it is made available, a conformity assessment procedure focusing on the substantially modified parts is required (Chapter 7).
Variant
Different components, configurations or enabled functionalities, such as one build per operating system or an edition with a larger feature set. A distinct product with its own placing on the market.
Only for standalone software
These timing rules apply exclusively insofar as the product is standalone software. Where software is combined with hardware, a different logic applies: software that is necessary to operate a device in accordance with its intended purpose is placed on the market at the same point in time as the hardware units (Chapter 4). Under the Regulation, manufacturers may keep older versions accessible in public software archives; users must then be informed clearly and in an easily accessible manner about the risks of unsupported software.
Software, web applications, websites and source code: where the product begins
Not all software is a product with digital elements. The guidance draws the line according to where software executes: what runs on the user’s side is a product; what is merely accessed remotely is not, on that basis alone. For source code, it separates two questions — whether code is software, and whether passing it on constitutes placing on the market.
Four types of product
The Regulation defines a product with digital elements as a software or hardware product and its remote data processing solutions, including software or hardware components placed on the market separately. It is covered where its intended purpose or reasonably foreseeable use includes a direct or indirect logical or physical data connection to a device or network. From this, the guidance derives four types:
- Standalone software, such as apps and computer programs, whether digitally or physically distributed.
- Hardware with embedded software, such as Internet-of-Things devices, laptops and tablets.
- Standalone hardware, such as integrated circuits and motherboards.
- Any combination of hardware and software supplied separately but intended to operate together.
The criterion: where does the software run?
Standalone software falls under the Regulation where it is placed on the market as a product. According to the guidance, it must be distinguished from the remote provision of services, which does not fall within the scope. A software product exists where the software is provided to a user, obtained by that user and operated on, or as part of, an electronic information system on the user’s side. This applies to software that is downloaded, installed or otherwise supplied and executes on the user’s side — expressly including browser extensions and applications developed using web technologies but supplied for local execution.
By contrast, software that executes remotely and is merely accessed by the user is not, on that basis alone, a product with digital elements. Under the Regulation, processing at a distance is covered only to the extent that it is necessary for a product to perform its functions — as remote data processing, not as a product in its own right. According to the guidance, this typically concerns web applications, including progressive web apps, that are accessed exclusively through a web browser. Websites may run on the user’s device to a limited extent, but they are not themselves products; they are covered only where they support the functionality of a product and thus qualify as remote data processing.
| Example | Scenario | Commission’s classification |
|---|---|---|
| Example 3 | Mobile app from an app store, installed on a smartphone | Product with digital elements; where placed on the market in the course of a commercial activity, it may fall within the scope |
| Example 4 | Desktop application built using web technologies and packaged for local installation | Product with digital elements; in the course of a commercial activity, it may be covered |
| Example 5 | Web application accessed only through the browser | Not a product; not covered unless it supports the functionality of a product |
| Example 5 | The same application as a locally installed client | Product with digital elements; where the client relies on remote data processing to perform functions, that processing is also part of the product |
| Example 6 | Website that merely presents information to visitors | Not a product; not in scope |
For apps and clients, the Commission uses cautious wording: they may be covered where they are placed on the market in the course of a commercial activity. In the official German language version of Example 5, the second paragraph says “Kunde” (customer); what is meant is the client.
Source code: software, yes — placing on the market, not always
The Regulation defines software as the part of an electronic information system which consists of computer code. According to the guidance, computer code covers both machine code, i.e. binary instructions executed directly by the processor, and source code, which is written in a programming language and must be compiled or interpreted. Whether a given supply of code amounts to placing on the market is a separate question. It requires a supply in the course of a commercial activity, and this is often not the case for source code. The guidance names three cases without placing on the market:
- Anyone who shares free and open-source software in publicly accessible repositories is generally not placing it on the market (Chapter 6).
- Unfinished code shared during the design and development phase, for example for testing or review, is not considered placed on the market, because the manufacturing phase is not complete.
- Sample or demo code in tutorials and training materials is likewise not considered placed on the market.
In addition, the guidance recalls that the Regulation allows unfinished software, such as alpha versions, beta versions or release candidates, to be made available on the market even if it does not comply, as long as it is made available only for the period necessary for testing and feedback. Where, by contrast, a manufacturer provides its customers with computer code as a product, that code is considered placed on the market — regardless of whether it is machine code or source code. In Example 7 of the guidance, company A licenses to company B the source code of a customisable internal platform, provided as a text file. Although B still has to adapt and compile the code, A is placing it on the market and is subject to the Regulation.
According to the English version, A is expressly not responsible for B’s subsequent adaptations and compilation. The official German language version published in the Official Journal says the opposite at this point; we consider this a translation error and follow the English version. A side note in the guidance concerns liability: for the purposes of product liability law, computer code as such is not a product, so manufacturers placing it on the market cannot be held liable under the Product Liability Directive (EU) 2024/2853. This does not rule out liability under national tort law.
Hardware with companion software, data connection, complex systems
For connected devices, it is the function, not the distribution channel, that determines what belongs to the product: drivers and companion apps without which a device cannot fulfil its purpose are part of it. Whether a device is covered at all depends on the data connection, not on the electronics. And complex systems with long life cycles are not exempt, but can demonstrate their conformity on a risk basis.
Companion software belongs to the device
According to the guidance, whether software forms part of a product with digital elements is determined not by how or when it is delivered. What matters is whether, in light of the intended purpose and reasonably foreseeable use, it is necessary for the product to perform its intended functions. Where a hardware product is designed to work together with a specific software from the same manufacturer, the two together constitute the product. Software that is necessary to operate, configure, control or use the product in accordance with its intended purpose is part of it — even if it is only obtained after the hardware has been placed on the market, via an app store, a download link or another digital channel. It is placed on the market at the same point in time as the hardware units; the timing rules for standalone software from Chapter 2 do not apply here.
Example 8: network printer
The printer is placed on the market as a hardware product, while the drivers for sending print jobs, configuring the device and managing its operation are available for download from the manufacturer’s website. Although they are supplied through different channels, the printer and the drivers form a single product: without the drivers, the printer cannot fulfil its intended purpose.
Example 9: fitness wearable
The wearable measures heart rate and activity. Measurements, history and configuration are available only via the manufacturer’s smartphone app, which is downloaded separately from the app store. The wearable and the app are designed and intended to operate together — and are therefore a single product.
Our assessment: for many manufacturers, this shifts the product boundary. App releases, driver updates and configuration tools are often created by different teams and in different release cycles from the firmware. For the purposes of the Regulation, they belong to the same product, with a joint risk assessment, technical documentation and conformity assessment. The case expressly addressed is software from the same manufacturer; the guidance does not expressly say whether necessary software from a third party is included. That remains a matter for case-by-case assessment.
Data connection: digital information rather than mere electronics
The Regulation applies to products whose intended purpose or reasonably foreseeable use includes a direct or indirect logical or physical data connection to a device or network. The guidance emphasises that the scope is anchored not in the mere presence of electronics, but in the capacity to exchange digital information. Because the Regulation does not define “data connection” or “digital data”, the Commission interprets the term. The aim is to distinguish products that merely use electrical signals from those that participate in digital communication and are therefore exposed to cybersecurity risks.
- 1Is information transmitted in binary form?
At its most basic level, a data connection involves transmitting information as a sequence of 0s and 1s.
- 2Are the states deliberately encoded as information?
There must be a sender that deliberately generates digital symbols according to a defined scheme. Simply switching an output on and off does not by itself constitute a data connection if those states are not intended to represent data or are not read by a digital input.
- 3Can the receiver interpret them as data?
At the destination, a receiver must be capable of decoding the symbols as information.
- 4Are the signals used solely to trigger or power a function?
Where electrical or electronic signals are used solely for this purpose, without conveying digitally encoded information, no data connection exists. To that extent, the product does not fall within the scope.
- 5Is the connection part of the intended purpose?
If a data connection exists, it remains to be checked whether it forms part of the intended purpose or reasonably foreseeable use — direct or indirect, logical or physical, to a device or network.
The criteria are taken from the guidance; the order of the steps is our own structuring. For borderline cases, you should document the outcome together with the reasoning.
Complex systems: not exempt, but risk-based
Products with digital elements can also be complex systems made up of multiple hardware and software elements operating together. Where such a system is placed on the market as a single product, it is a product within the meaning of the Regulation. The guidance describes typical characteristics: long design and development cycles, contracts that may have been signed before the date of application, particularly long operational lifetimes, a high degree of technical and organisational complexity, components placed on the market before the date of application, established architectures and widely used interoperability standards. Certain technical characteristics may then be difficult or disproportionate to modify without affecting the intended purpose, safety, reliability or interoperability with existing infrastructure.
These characteristics do not exclude such systems from the scope. According to the guidance, they illustrate the Regulation’s risk-based approach, which allows compliance to be demonstrated in different ways. Where certain essential requirements are not applicable or cannot be fulfilled with state-of-the-art measures — for instance because the system must interact with existing dependencies or meet interoperability requirements — manufacturers should identify and document these constraints, assess the risks and implement alternative or compensatory risk-mitigation measures. The technical documentation and the user information describe the constraints, risks and measures transparently. During the support period, manufacturers must periodically reassess whether the constraints continue to exist; where they can be lifted or reduced, the product should be updated. According to the guidance, this logic may apply, as appropriate, to all products, not just complex systems.
| Example | Situation | Commission’s expectation |
|---|---|---|
| Example 10: legacy protocol | The risk assessment requires a secure communication protocol, but the intended purpose includes interoperability with existing systems that only support an older, less secure protocol. | The less secure protocol is permissible insofar as it is necessary for interoperability and its risks are identified and mitigated by other means. Where supporting both is technically feasible, the secure protocol is implemented and used by default. |
| Example 11: legacy component | A manufacturer of industrial automation systems purchased microcontrollers before the date of application; the system is to be placed on the market afterwards. The risk assessment identifies no specific risks for one of these microcontrollers. | The manufacturer integrates it and ensures that the system as a whole meets the essential requirements. |
Legacy designs: products designed before the date of application on 11 December 2027
Many products placed on the market from 11 December 2027 onwards were designed years earlier. The guidance clarifies that the Regulation does not automatically require them to be redesigned: the risk assessment decides. Conformity assessment, the EU declaration of conformity, CE marking and vulnerability handling nevertheless remain mandatory.
New units of an old type
From 11 December 2027, the Regulation applies in full. After that date, many manufacturers will place on the market units manufactured in accordance with a type or model designed and developed before that date. Because placing on the market refers to each individual product (Chapter 2), the obligations of the Regulation apply to these units as to any other product. The guidance answers the question of what this means for the design: compliance with the Regulation does not necessarily require these units to be redesigned.
Terminology trap in the official German language version: in this section, it refers to the Regulation’s “Inkrafttreten” (entry into force) on 11 December 2027. What is meant is the date of application; the Regulation already entered into force on 10 December 2024. We therefore use “date of application” throughout.
The risk assessment decides
The manufacturer must carry out a cybersecurity risk assessment. On the basis of the intended purpose and reasonably foreseeable use, this assessment determines which essential cybersecurity requirements are applicable to the product and how they are implemented. Where the assessment shows that the product already incorporates appropriate and effective security measures addressing the relevant risks, the manufacturer may rely on those measures to demonstrate compliance. The Regulation does not in itself impose an obligation to introduce new security features or to redesign the product where this is not necessary to address the identified risks. Conversely, in our reading, this means that if the assessment reveals gaps, new measures are needed to the extent that the risks require them.
Where the manufacturer cannot demonstrate how the risk assessment was taken into account during the original design and development phase, the guidance states that the obligation is to be understood as follows: the manufacturer carries out the assessment now and demonstrates on that basis that the product incorporates adequate security measures with a view to minimising cybersecurity risks, preventing incidents and minimising their impact — including in relation to the health and safety of users. Where the risk assessment and technical documentation demonstrate an appropriate level of cybersecurity and compliance with the essential requirements, the product can be placed on the market without redesign.
Proportionate evidence, obligations in full
The manufacturer demonstrates conformity through the relevant conformity assessment procedure and includes the evidence in the technical documentation. The guidance interprets this obligation in light of the risk assessment. For legacy designs subject to the procedures of the so-called “default” category, i.e. without a core functionality of a category of important or critical products (Chapter 9), no test results from the original design and development phase should be required where the risk assessment already demonstrates appropriate and effective measures. Such evidence would not increase the security of the product. Where tests are nonetheless necessary, evidence is not expected for every product variant; the tests can be grouped across product families.
All other obligations remain, regardless of whether the design had to be modified. Before the product is placed on the market, the applicable conformity assessment procedure must have been carried out, the EU declaration of conformity drawn up and the CE marking affixed. The manufacturer should provide evidence of how it meets the vulnerability handling requirements, keep the risk assessment up to date and provide users with the required information and instructions.
Example 12 of the guidance: a microcontroller from before the date of application
- 1Starting point
A microcontroller designed and developed before the date of application is intended to be integrated into a range of electronic products, including products with connectivity functionalities. New units are to be placed on the market after the date of application.
- 2Risk assessment
Beforehand, the manufacturer assesses the cybersecurity risks on the basis of the intended purpose and reasonably foreseeable use. Among other things, it identifies unauthorised access, manipulation of software or data, and misuse of available interfaces.
- 3Outcome
The original design already incorporates appropriate and effective measures against these risks. The manufacturer concludes that the relevant essential requirements are met; no redesign or additional security functionalities are necessary.
- 4Evidence
Where an earlier risk assessment cannot be evidenced, the manufacturer documents a current assessment and explains how the design and measures mitigate the risks. It is not required to recreate historical design or test documentation. For variants with the same design and the same risk profile, representative evidence for the product family is sufficient.
- 5Ongoing obligations
The manufacturer maintains processes to address and remediate vulnerabilities and keeps the risk assessment under continuous review.
- 6Placing on the market without redesign
In these circumstances, the microcontroller may be placed on the market without redesign, provided that the risk assessment and technical documentation demonstrate an appropriate level of cybersecurity and the applicable essential requirements are met.
Legacy design, existing product, legacy component
| Scenario | What applies | Chapter |
|---|---|---|
| Legacy design: type designed before 11 December 2027, unit placed on the market afterwards | All manufacturer obligations for the unit; redesign only to the extent that the risk assessment requires it | 5 |
| Existing product: unit placed on the market before 11 December 2027 | The law applicable at the time of placing on the market is decisive, as long as no substantial modification takes place on or after 11 December 2027; the reporting obligations have nevertheless applied since 11 September 2026 | 7 and 12 |
| Legacy component: procured before the date of application, integrated into a product placed on the market afterwards | Part of the risk assessment of the overall product; the product as a whole must meet the essential requirements (Example 11) | 4 |
The overview summarises statements from several sections of the guidance; the terms legacy design, existing product and legacy component are our own.
Free and open-source software: responsibility, monetisation, stewards
Open source is neither exempt across the board nor covered across the board. The guidance proceeds step by step: is it free and open-source software within the meaning of the Regulation, who bears responsibility, and is it placed on the market in the course of a commercial activity? A legal person that is not a manufacturer can be an open-source software steward — with considerably lighter obligations.
The Regulation applies to products made available for the first time in the course of a commercial activity. According to the “Blue Guide”, whether a commercial activity exists can only be assessed on a case-by-case basis, for example taking into account the regularity of the supplies, the characteristics of the product and the intentions of the supplier. For free and open-source software, the Regulation sets its own guard rails: the circumstances of development and the way it is financed should play no role, and the provision of non-monetised free and open-source software is not considered a commercial activity. If such software is not placed on the market, it falls outside the scope — unless the entity publishing it is a legal person that meets the definition of an open-source software steward. In that case, the entity is subject only to the obligations of stewards.
The assessment sequence in eight steps
- 1Is it free and open-source software within the meaning of the Regulation?
Two conditions must be met cumulatively: a licence granting all rights to make the software freely accessible, usable, modifiable and redistributable, and openly shared source code. Openly shared means publicly available, upstream or downstream. Software under a free licence whose source code is available only to paying customers or a limited group does not meet the definition. The guidance does not name specific licences.
- 2Is the software under your responsibility?Example 13
Responsibility lies with whoever publishes it and exercises primary control over its development, releases and distribution decisions — in practice, the maintainers. Anyone who only contributes code and does not control releases, the roadmap or governance is a contributor; to that extent, the Regulation does not apply to them. Commit access alone does not establish responsibility. Anyone who submits a pull or merge request is likewise a contributor.
- 3Do you charge a price for the software itself?
Charging a price, for example for pre-compiled binaries, constitutes placing on the market; whoever supplies software on that basis is a manufacturer. Where a paid or enhanced version exists alongside a free community version, including in the open-core model, these are different products: the monetised version is placed on the market, the community version is not. For the community version, a legal person is subject to the steward obligations; a natural person is subject to none.
- 4Do you monetise other services or data through the software?Examples 14–16
This, too, constitutes placing on the market: where other products or services are monetised through the software, for example through advertising, commission fees, subscriptions or paid additional servers, or where use is conditional on the processing of personal data for purposes other than exclusively security, compatibility or interoperability.
- 5Is access tied to remuneration or donations?Examples 17–22
Paid consultancy, training or installation assistance around freely available software does not in itself constitute placing on the market. The decisive factor is whether access to the software itself, including its maintenance, is conditional on remuneration, for example in the case of a paid edition with technical assistance. A donation link is not to be viewed as an intention to make a profit, even where donations exceed the costs. The situation is different where releases, binaries, updates or security fixes are de facto available only to donors, or where donations bring contractual benefits or exclusive advantages that go beyond community perks.
- 6Financing, not-for-profit status, integration by othersExamples 23–26
Grants, bug bounty programmes, sponsorships or paid feature development do not turn openly shared software that is not otherwise monetised into a product placed on the market. Software published by not-for-profit entities whose earnings after costs are all used to achieve not-for-profit objectives is not considered placed on the market, even where it is directly monetised. The same applies to software intended for integration by other manufacturers that the publisher does not additionally monetise.
- 7Are you an open-source software steward?
If the software is not placed on the market, a legal person checks whether it is a steward: this is the case where the software is intended for commercial activities, for example for integration into commercial services or monetised products, and the legal person systematically provides support on a sustained basis for its development and ensures its viability. Natural persons cannot be stewards; without placing on the market, they have no obligations under the Regulation.
- 8Do you integrate third-party free and open-source software?Examples 27–34
In that case, all manufacturer obligations apply to your own product, including due diligence for each integrated component. You do not become responsible for the compliance of the component itself, not even through code contributions, developer time or funding.
The steps follow the stylised flowchart in the guidance and supplement it with criteria from the text; the order is our own structuring. The assessment is made per project, per version and per edition. For donations, the guidance requires an overall assessment of the circumstances.
Monetisation: the borderline cases
| Scenario | Example | Outcome under the guidance |
|---|---|---|
| Free operating system, alongside a paid version with technical assistance and performance optimisation | 17 | The paid version is placed on the market |
| Freely downloadable command line tool, consultancy and training can be booked separately | 18 | Not placed on the market |
| Service provider installs third-party software at a customer’s site without substantially modifying it | 19 | No placing on the market by the service provider |
| Call for donations; access to software, source code and updates independent of donations | 20 | Not placed on the market |
| Downloadable releases and security updates only for donors | 21 | Placed on the market: the donation amounts to charging a price |
| Source code public; binaries, updates and guaranteed security fixes only for donors | 22 | Placed on the market |
| Manufacturer pays an individual developer for a feature that is openly shared | 23 | Not placed on the market; if the manufacturer integrates the software, its due diligence obligation applies |
| Not-for-profit publisher of a browser, monetised via search engine partnerships | 24 | Not placed on the market; the publisher is a steward |
Paid enterprise versions under a commercial agreement are considered placed on the market, even where functionally equivalent software is available free of charge under a free licence. For natural persons, directly bundled technical assistance remains unproblematic as long as the price covers only the actual costs, including reasonable living expenses.
Open-source software stewards: obligations by type of support
With the open-source software steward, the Regulation has created a legal category of its own: a legal person, other than a manufacturer, that has the purpose or objective of systematically providing support on a sustained basis for the development of free and open-source software intended for commercial activities, and of ensuring the viability of that software. Typical forms of sustained support are hosting and managing collaboration platforms, hosting source code or software, governing or managing the software and steering its development. The status applies per project: the same legal person can be a steward for one project and a manufacturer for another, a steward for the community version and a manufacturer for the paid version of the same software. Foundations are stewards only for those projects that they systematically support and that are intended for commercial activities.
| Type of support | Typical activities | Reporting and information obligations under the guidance |
|---|---|---|
| Non-technical only | Branding, governance rules, community events, collecting donations | No obligation to report actively exploited vulnerabilities or severe incidents. If the steward becomes aware of an actively exploited vulnerability, it should share the information with the maintainers in accordance with its cybersecurity policy; voluntary reporting should be considered. |
| IT infrastructure | Hosting source code repositories, version control systems, generating signing keys | Notify ENISA and the CSIRTs of severe incidents related to this infrastructure that have an impact on the security of products; where appropriate, inform all users, for example via a general announcement. No obligation to report actively exploited vulnerabilities, but voluntary reporting should be considered. |
| Engineering resources | Employing developers, coordinating development, reviewing and merging code, managing releases, handling vulnerability reports and security patches | Report actively exploited vulnerabilities of which the steward becomes aware and, where appropriate, inform all users; where there is a direct relationship with impacted users, inform them directly. |
All stewards must comply with the Regulation’s general steward obligations; how far the reporting and information obligations extend depends on the type of support.
Where stewards must report actively exploited vulnerabilities, the guidance states that the obligation arises as soon as active exploitation is detected — not merely because a vulnerability exists in the codebase — and only to the extent that the steward is involved in development. In practice, the steward is likely to learn of this through reports from third parties, such as manufacturers, users or security researchers (Chapter 12). The status can change: where an entity ceases to provide systematic support on a sustained basis for a specific piece of software, it may no longer meet the definition; the guidance encourages it to communicate the change in status clearly. If a steward begins to monetise the project directly, it places it on the market and is a manufacturer from that point on — not retroactively for the versions it supported as a steward.
Integration by other manufacturers
For manufacturers that build free and open-source components into their products, three statements are central. First, whether the Regulation applies to a component depends solely on whether its publisher places it on the market; integration into a commercial product does not change its status, and maintainers of components not placed on the market have no obligations towards integrators. Second, whoever integrates a component does not become responsible for that individual component’s compliance — not even by contributing code to its maintenance. Third, all manufacturer obligations apply to the integrator’s own product: due diligence when integrating each component, and the obligation to report vulnerabilities in integrated components upstream and to share security patches. The scenarios in Examples 27 to 34, which the guidance describes as purely hypothetical, play this through.
Individual developer with a donation link (Examples 27 and 34)
Software openly shared, no price charged, voluntary donations, including from companies that integrate it; in Example 34, the library is published on a public package repository. Not placed on the market: the developer has no obligations under the Regulation, and in the scenario of Example 34 neither does the package repository. The integrating manufacturers bear their due diligence obligation.
Not-for-profit foundation or entity (Examples 28, 32, 33)
Component or software development kit for integration into commercial products, supported on a sustained basis and financed through public funding, donations, partnership projects or membership fees. The entity is the steward. Contributing manufacturers and members do not become manufacturers of the component and are not responsible for its compliance.
Company without monetisation (Examples 29 and 31)
A company publishes a component under its own name and maintains it so that others can integrate it, without charging a price and without any other form of monetisation. It is a steward, not a manufacturer. Anyone who only contributes code and developer time and offers support independently has no obligations for this software.
Company with a paid version (Example 30)
A company offers its software as a paid version with technical assistance or performance optimisation; other companies contribute, but control remains with the publisher. The publisher is a manufacturer; as a not-for-profit entity, it would be a steward. The contributors have no obligations for this software; if they integrate it, their due diligence obligation applies.
Substantial modification: repairs, spare parts, software updates
Whether a modification is substantial determines whether a new conformity assessment and a new placing on the market are required and, in some circumstances, whether there is a new manufacturer. The guidance measures this not by scale or effort, but by risk: new or increased risks not covered by the risk assessment, or a changed intended purpose, make a modification substantial. Repairs and security updates are generally not substantial modifications.
Definition and significance
Under the Regulation, a substantial modification is a change to the product following its placing on the market which affects compliance with the essential cybersecurity requirements or results in a modification to the intended purpose for which the product has been assessed. The concept determines obligations in four situations. Importers and distributors that substantially modify a product already placed on the market are considered manufacturers, as is any other person who substantially modifies a product and makes it available on the market. Anyone who, after 11 December 2027, substantially modifies a product placed on the market before that date and places it on the market is considered its manufacturer. And for products placed on the market after that date, the concept determines whether a later change requires a new conformity assessment procedure — particularly for iteratively developed software.
Repairs and spare parts
According to the recitals of the Regulation, refurbishment, maintenance and repair do not necessarily lead to a substantial modification, for instance where the intended purpose and functionalities are not changed and the level of risk remains unaffected; an upgrade by the manufacturer, by contrast, may affect design, intended purpose and compliance. The guidance requires a case-by-case assessment for physical modifications and states that such operations are generally not substantial modifications. Nor does replacing defective or worn parts with better-performing ones constitute a substantial modification per se. It only does so if the change in performance or in the way the product operates affects compliance with the essential requirements or changes the intended purpose in a way not covered by the risk assessment. In Example 35, a server manufacturer replaces defective RAM with better-performing RAM; the higher performance remains within the assessed use, and the server is not substantially modified.
For spare parts, the Regulation provides for an exemption: parts intended to replace identical components and manufactured according to the same specifications are not subject to it. The guidance draws two limits. First, the exemption applies only where the part is specifically supplied to repair or extend the durability of a product already placed on the market, whether that product was placed on the market before or after 11 December 2027. This must be apparent from the context of the supply, for example through identification of the product or product family in the order or commercial offer, or through supply via after-sales and service channels; supporting evidence should be kept available to the market surveillance authorities. The fact that a part technically fits is not sufficient. Second, whether a part is “identical” is measured by its functional role and the characteristics relevant to cybersecurity: differences that do not affect the security characteristics or the risk profile are not detrimental, whereas differences in algorithms, protocols, cryptographic mechanisms or access controls may rule out identity. The exemption also applies where a complete product that forms part of a larger product is replaced.
| Example | Spare part | Outcome |
|---|---|---|
| 36 | Identical communication module for connected controllers placed on the market in 2026, before the date of application, and in 2028, in compliance with the Regulation | Exemption applies in both cases; the repair is not a substantial modification |
| 37 | Newer communication chip with a different cryptographic implementation and updated secure boot for an industrial controller from 2026 | Not identical: a product in its own right that must comply with the Regulation, assessed also with regard to interoperability with the legacy device |
| 38 | Wireless module with a different chipset and updated firmware, but the same protocols and security mechanisms | Can be considered identical; exemption applies |
| 39 | Replacement CPU or complete replacement PLC for an automation system, supplied via the after-sales service channel with the target system identified | Exemption applies in both variants |
If a spare part is not identical, it is a product in its own right. Where individual essential requirements cannot be met because of its intended purpose or technical constraints, the same applies as for complex systems: reflect this in the risk assessment, implement alternative or compensatory measures and document them transparently (Chapter 4). This does not in itself make the repair of the main product a substantial modification.
Software updates: risk, not scale
The guidance devotes the most space to software updates. The starting point is the recitals of the Regulation: a product is substantially modified where a change alters the level of cybersecurity risk and this altered or additional risk was not considered in the risk assessment and, consequently, in the implementation of the essential requirements. From this, the Commission derives a case-by-case assessment for each update. New functionalities that change the intended purpose of the product as a whole are generally substantial. Progressively introduced functionalities that the risk assessment has already anticipated and assessed, or that do not, in themselves, alter the risk profile, are not. Conversely, even small functionalities can introduce significant risks: the yardstick is not the scale or complexity of the change, but its potential adverse impact on the risk profile. Security updates are generally not substantial, even where they introduce significant technical changes or constrain functionalities solely to mitigate vulnerabilities — unless they change the intended purpose or introduce new risks, for example by materially changing the product’s boundaries or dependencies.
| Update | Example | Substantial? | Why |
|---|---|---|---|
| Dashboard can now control machines instead of merely displaying data | 40 | Yes | Intended purpose shifts from situational awareness to operational control, not envisaged in the risk assessment |
| Consumer app generates behavioural profiles and makes automated decisions about content | 41 | Yes | The user-controlled tool becomes an automated decision-making system |
| Group chat in a messaging app, foreseen in the design and the risk assessment | 42 | No | Remains within the assessed intended purpose |
| Activation of already built-in, assessed control loops, including safeguards | 43 | No | Implemented as assessed |
| “Remember me” function with locally stored tokens | 44 | Yes | Despite its limited scope, new risks not assessed, such as token theft and session hijacking |
| Export of detailed system logs, stored unencrypted | 45 | Yes | Risks of data exposure that were neither assessed nor mitigated |
| Patch against a buffer overflow or authentication bypass | 46 | No | Only the internal implementation changes; the update serves exclusively to reduce risk |
| Hardening: stricter firewall rules, closing unused ports, password policies, mandatory MFA where foreseen | 47 | No | Purpose unchanged, even if users configure or access the product differently |
| Disabling a deprecated algorithm, activating a stronger, already supported alternative | 48 | No | Foreseen and assessed, without new external dependencies or altered data flows |
| Local encryption replaced by mandatory upload to a remote encryption service | 49 | Yes | Intended purpose fundamentally altered, although security-driven |
| Key management switched to an external service operated by a third party | 50 | Yes | New dependencies, data flows and external interfaces, not assessed |
Summary of Examples 40 to 50 of the guidance. The outcomes illustrate the Commission’s interpretation and do not replace a case-by-case assessment.
Four questions for every update — non-exhaustive
- 1New threat vectors?
Additional interfaces, communication channels, execution environments or external dependencies.
- 2New attack scenarios?
New ways in which unauthorised access, manipulation, interference or misuse of the product, or of the data it processes, could occur.
- 3Changed likelihood of known attacks?
Less effort or expertise required to carry them out, greater exposure to untrusted actors, weakened safeguards.
- 4Changed impact of known attacks?
More data or functions affected, more severe operational, safety or economic consequences, reduced ability to detect, contain and recover from an incident.
- 5Outcome
If none of the questions applies and the assumptions and mitigation measures of the risk assessment remain valid and effective, the update is likely not a substantial modification. If one applies, the manufacturer reassesses the risks and checks whether the essential requirements continue to be met and whether new or increased risks arise that were not foreseen.
- 6Always, regardless of the outcome
During the support period, manufacturers remain responsible for the security of their updates and of the product in accordance with the vulnerability handling requirements. The risk assessment and technical documentation must be kept accurate, complete and continuously up to date, including for updates that are not a substantial modification.
Consequences: new product, new manufacturer?
If a substantially modified product is made available on the market, it is a new product for the purposes of the Regulation; making it available constitutes a new placing on the market, regardless of who carried out the modification. If the original manufacturer makes the modification, including in the course of iterative software development, it remains the manufacturer. It may reuse existing documentation and test results for parts not affected, and the conformity assessment procedure should focus on the substantially modified parts — as should the conformity assessment body in the case of a third-party assessment. By contrast, anyone who substantially modifies another manufacturer’s product and makes it available changes role.
| Who modifies? | Role afterwards | Scope of obligations |
|---|---|---|
| Original manufacturer | Remains the manufacturer; the modified product is considered newly placed on the market | Conformity assessment focusing on the modified parts; reuse documentation and tests for unchanged parts |
| Importer, distributor or other person making the modified product available | Considered the manufacturer of the modified product, even without involvement in the design or the original placing on the market | Manufacturer obligations, including conformity assessment and reporting obligations, only for the modified part, as long as the cybersecurity of the product as a whole is not negatively affected; otherwise for the entire product |
| Integrator assembling components into a product of its own | Manufacturer of a new product; not a substantial modification of another manufacturer’s product | The Regulation applies in full to the product as a whole; the integrator may rely on the compliance activities of the component manufacturers (Example 51) |
| Original manufacturer of a product placed on the market before 11 December 2027 | Remains the manufacturer | Only the substantially modified parts, unless the modification negatively affects the cybersecurity of the product as a whole |
Where a third party makes the modification, the original manufacturer’s obligations continue to apply to the unchanged original product. According to the “Blue Guide” cited in the guidance, whoever makes the modification must demonstrate that not all parts of the technical documentation need to be updated, and draws up their own declaration of conformity.
Support period: expected use time instead of a blanket five years
As a manufacturer, you decide yourself how long you handle a product’s vulnerabilities — but not at will. The guidance anchors the support period in the expected use time, reads the Regulation’s five years as a minimum and clarifies what new versions and substantial modifications change about the period. The reporting obligations do not end with it.
The Regulation requires every manufacturer to determine a support period. During this period, vulnerabilities of the product, expressly including its components, must be handled effectively and in accordance with the vulnerability handling requirements. The period is therefore the framework for your entire vulnerability management on the market: for security updates, for effective and regular tests and for the disclosure of fixed vulnerabilities. Because the Regulation expressly calls for guidance on this concept, the Commission devotes a separate section to it.
The yardstick: expected use time
The support period is meant to reflect how long the product is expected to be in use. To this end, the Regulation names in particular reasonable user expectations, the nature of the product, including its intended purpose, and relevant Union law determining the lifetime of products. The guidance lists further criteria that you may take into account and requires all criteria to be applied proportionately. Our assessment: the period is a reasoned manufacturer decision, not a marketing claim — and the reasoning should be documented in such a way that a market surveillance authority can follow it.
User expectations
How long can buyers reasonably expect to use the product securely? For a programmable logic controller, the answer differs from that for a short-lived consumer app.
Nature of the product and intended purpose
Durability, wear and the deployment environment shape the use time. The examples in the guidance repeatedly rely on the physical durability of the hardware.
Union law on lifetime
Where other Union legislation determines a lifetime for your product, that lifetime feeds into the determination of the support period.
Further criteria
Support periods of comparable products from other manufacturers, the availability of the operating environment, the support periods of components sourced from third parties that provide core functions, and relevant guidance from the administrative cooperation group (ADCO) and the Commission.
Five years is a minimum, not a default
The Regulation prescribes at least five years. The period may only be shorter if the product is expected to be in use for less than five years; it then corresponds to the expected use time. The guidance expressly describes this minimum duration as a safeguard: it is intended to ensure that vulnerabilities are handled for a sufficiently long period, while leaving room for products with genuinely shorter use. Where use for more than five years can reasonably be expected, support periods should be correspondingly longer. The Commission therefore clarifies that five years is not to be regarded as the default for all products. For you, this means: a period of less than five years requires evidence of the shorter use; where longer use is expected, a blanket five-year statement is not sufficient.
State the end date, notify the expiry
The Regulation requires transparency. At the time of purchase, the end date of the support period must be indicated in a clear and understandable manner, at least the month and year. At the end of the period, manufacturers must display a notification to users, where this is technically feasible in light of the nature of the product. According to the guidance, this is so that users know how long they can expect security support. VamiSec recommends showing the end date where the purchasing decision is made — on the product page, in the shop, in the quotation or in the contract — and feeding it from the same source as the technical documentation.
Software: every substantially modified version needs its own period
Software is often developed iteratively. The guidance therefore requires each substantially modified version to have a declared support period that complies with the Regulation’s criteria — including the five years, unless the use time of that version is demonstrably shorter. Where a substantial modification leaves the factors determining the use time unchanged, however, the remaining period applies; more on this below. At the same time, the Regulation offers flexibility: you may limit the remediation of vulnerabilities to the version last placed on the market if users of earlier versions have access to that version free of charge and do not incur additional costs to adjust their hardware and software environment. The Commission interprets “additional costs” in practical terms. Personnel time, routine testing, configuration adjustments or updating dependencies that have reached the end of their life do not count. Mandatory purchases of new hardware, infrastructure replacement or fundamental changes to the operating environment, by contrast, do.
Example 52 of the guidance illustrates this for a smartphone whose operating system can be updated to substantially modified versions free of charge and without new hardware. Example 53 concerns enterprise software released in a substantially modified version every few months. In both cases, the manufacturer may discontinue remediation for older versions as soon as an upgrade is possible, even if this shortens their effective support period. The other obligations continue to apply, such as the policy on coordinated vulnerability disclosure and the measures for sharing information about potential vulnerabilities, as do the reporting obligations. Users who have not upgraded should be informed, where technically feasible. If a free upgrade is possible, the Regulation does not oblige you to keep supporting older versions free of charge; you may, however, do so, including for a fee.
Substantial modification: reassessment, not a reset
If a product is substantially modified and made available again, it is treated as a new product with a new placing on the market. In the Commission’s view, the consequence for the support period is a reassessment against the same criteria, but neither an automatic reset nor an automatic extension. One question is decisive: does the modification affect the factors that originally determined the expected use time? If not, the period corresponds to the remaining original use time, even if less than five years remain. If it does, you recalculate it, for example after revitalising hardware modifications that extend the lifetime, or after a substantial reprogramming or rewrite of the core software.
| Example | Substantial modification | Does it affect the use-time factors? | Consequence for the support period |
|---|---|---|---|
| Example 54: robot vacuum cleaner | Software update adding new cleaning modes and navigation features | No: hardware durability and user expectations remain the same | Remaining expected use time as originally determined |
| Example 55: industrial machine with a cloud back-end | Rearchitecting the back-end, i.e. the remote data processing solution, with new APIs and data flows | No: the nature of the product and user expectations remain the same | Remaining expected use time as originally determined |
| Example 56: programmable logic controller | Replacement of the embedded computing platform, including processor, memory and runtime environment, with a longer-lasting generation | Yes: users may now expect a longer period of productive use | Recalculation based on the longer expected use time |
In these examples, the guidance does not state any numbers of years but works with placeholders such as “X years”. According to the guidance, repair, maintenance and refurbishment are generally not considered substantial modifications.
Important and critical products: the core functionality determines the procedure
Whether a product counts as a default product, an important product of class I or II or a critical product determines its route through conformity assessment. To this end, the guidance gives practical shape to a concept that the Regulation does not define: core functionality. It is determined objectively, there is exactly one, and integrated components do not shift it.
The Regulation contains two lists: important products, divided into classes I and II, and critical products. A product belongs to a category if it has that category’s core functionality; the technical description of the categories is set out in Implementing Regulation (EU) 2025/2392. For all other products, the guidance uses the term “default category”, a working term of the Commission that the Regulation itself does not use. The classification has consequences: it determines the conformity assessment procedure that must be completed before the product is placed on the market. Under the Regulation, breaches of the conformity assessment obligations can be sanctioned with administrative fines.
What core functionality is
The Commission understands core functionality as the main features and technical capabilities without which a product would not be able to meet its intended purpose. It is assessed in the specific context and under the conditions of use, based on, among other things, the instructions for use, promotional and sales materials, statements by the manufacturer and the technical documentation. What counts is the objective finding: how a product is described or marketed does not matter if the description does not reflect its actual technical characteristics. The guidance expressly prohibits presenting the core functionality, by overemphasising or downplaying individual functions, in such a way that the more stringent procedure is circumvented. Clear inconsistencies between promotional materials, instructions for use and technical documentation are a warning sign.
For the purpose of determining the procedure, a product has no more than one core functionality. It should be clearly identified in the technical documentation so that the applicable procedure is established and market surveillance authorities can check its application. Additional functions do not rule out a category: if they merely complement or enhance a core functionality that corresponds to a category, the classification stands. Modules that you supply solely as part of an integrated product are not classified individually; the core functionality is then determined at the level of the product as a whole.
Five cases on core functionality
| Example | Product | Outcome according to the guidance | What follows from it |
|---|---|---|---|
| Example 57 | Software that, as an abstraction layer over the hardware, manages processes, memory, input and output as well as resources and provides system interfaces | Core functionality of an operating system within the meaning of the category description | The technical documentation and the manufacturer’s statements support the classification |
| Example 58 | Smartphone with an integrated operating system | No operating system core functionality; the smartphone as a whole serves communication and access to information and services | The mere integration of an important or critical product does not in itself make the end product important or critical |
| Example 59 | SOAR software that can also perform SIEM functions | Generally no SIEM core functionality, because it substantially exceeds it, for example with incident response | A product that can perform a category’s functions does not automatically belong to it |
| Example 60 | Log collection and visualisation tool with basic dashboards | Generally no SIEM core functionality, because data correlation and actionable security insights are missing | A product that falls substantially short of a category does not belong to it either |
| Example 61 | Security suite with SIEM, intrusion detection and analytics modules, the modules also available as separate subscriptions | SIEM module class I, intrusion detection system class II, analytics module default | Separately offered modules are standalone products with their own core functionality |
Which procedure for which class
| Classification | Conformity assessment | Third-party involvement |
|---|---|---|
| Default | The internal control procedure based on module A (“self-assessment”) is always possible; you may voluntarily choose a more stringent procedure | Not required |
| Important, class I | Module A only if you have fully applied a relevant harmonised standard and its scope covers at least all cybersecurity risks of the core functionality; the same applies to common specifications and European certification schemes at assurance level at least “substantial” | Required if these bases have not been applied or have been applied only in part |
| Important, class II | Module B+C, module H or a cybersecurity certification scheme | Always required |
| Critical | Module B+C, module H or a cybersecurity certification scheme | Always required |
Exception under the Regulation: manufacturers of important products of classes I and II placed on the market as free and open-source software may follow the procedures of the default category; the Regulation makes this conditional on the technical documentation being publicly available at the time of placing on the market. According to the guidance, a European certification scheme at assurance level at least “substantial” counts as third-party involvement because a third-party certification body is involved.
For class I, the Commission sets out two cumulative conditions for the internal control procedure: all applicable requirements of a relevant harmonised standard have been applied, and the standard’s scope covers at least all cybersecurity risks of the core functionality. If your product goes beyond the scope of the standard, module A remains possible. You must then, however, demonstrate that the remaining risks are also addressed and document the additional measures; the risk assessment is mandatory in any case. Example 62 of the guidance shows this for antivirus software with additional disk-cleaning and anti-tracking functions, Example 63 for a router with an integrated firewall component. There, the manufacturer may additionally apply the harmonised standard for firewalls to the firewall functionality. In both cases, the internal control procedure is permissible for the product as a whole — according to the guidance, even where a product with the core functionality of a class I product contains a component that would in itself be a critical product.
For class II, apart from the open-source exception, and for critical products, this route does not exist; a third party is always involved. The Regulation also allows the Commission to make European cybersecurity certification mandatory for critical products. The guidance does not address any such determination.
Presumption of conformity only for covered risks
If you apply a harmonised standard whose reference has been published in the Official Journal, you benefit from the presumption of conformity — but only for the risks and essential cybersecurity requirements covered by that standard. It can apply to the product as a whole, for example, where the standard covers all risks of the core functionality and no additional function brings its own uncovered risks. The Commission expects, for instance, that the harmonised standard for routers will cover at least the risks of the router core functionality. Example 64 of the guidance: antivirus software may be assessed under the internal control procedure, but the presumption applies only to the core functionality, not to disk cleaning and anti-tracking. Example 65: if the standard is extended to cover disk cleaning, the presumption extends to it; the anti-tracking function remains outside. Irrespective of this, you bear full responsibility for assessing all risks of your product.
For context: Examples 62 to 65 presuppose harmonised standards whose references have been published in the Official Journal. Whether a class I product can use the internal control procedure therefore depends on whether such a standard or an equivalent basis exists for its category. Our recommendation: plan for both routes — module A once references of harmonised standards have been published in the Official Journal, and third-party conformity assessment for as long as this is not the case.
Risk assessment, components and product families: the product is the yardstick
Section 7 of the guidance sharpens the obligation with which everything begins: the cybersecurity risk assessment. The Commission measures residual risks not against your risk appetite but against the appropriate level of cybersecurity of the product, and responsibility cannot be handed off by means of a warning notice. A separate due diligence obligation applies to components sourced from third parties, and pragmatic reuse of documentation applies to product families.
The Regulation establishes two distinct but complementary obligations. First, manufacturers assess the cybersecurity risks of the product itself: they identify relevant risks, assess their potential impact and implement appropriate measures in line with the essential cybersecurity requirements. Second, a due diligence obligation applies to integrated components. Together, the two are intended to ensure that the product remains secure, including where it relies on remote data processing solutions, remote services or third-party software and hardware components.
Residual risk: the yardstick is the product, not your risk appetite
In organisational risk management, decisions are based on acceptance criteria derived from corporate objectives or risk appetite. The guidance deliberately sets a different yardstick: residual risk is to be measured by whether the product ensures an appropriate level of cybersecurity based on the risks, taking into account its intended purpose and reasonably foreseeable use. In the Commission’s view, internal risk tolerance, commercial strategy or mere cost considerations are not relevant to this. The assessment must take into account the conditions of use, including the operational environment and the intended users, the assets to be protected and the expected use time. The Commission recognises that residual risks cannot be entirely eliminated. This does not, however, imply any discretion to accept arbitrary risks: a product may only be placed on the market once the residual risks have been sufficiently addressed through the implementation of the essential requirements. Where risks cannot be adequately addressed, changes to the design, functionality or intended purpose may become necessary; cost or commercial feasibility alone does not justify leaving them untreated.
Address risks instead of shifting them
The Regulation requires manufacturers to address identified risks through appropriate measures, including at product level. The guidance gives as examples a reduced attack surface, technical safeguards, limited or adapted functionality and a more precisely defined intended purpose. Foreseeable use can also be shaped, through risk communication, user guidance or a user interface that steers towards secure use patterns. The guidance draws a clear line: the Regulation does not provide for the transfer of cybersecurity risks or responsibility to users or third parties to compensate for design shortcomings. Information and instructions may support secure deployment and inform users about residual risks, in a manner appropriate to the type of product and its intended users.
Two examples show the latitude. In Example 66 of the guidance, a manufacturer refrains from protecting an industrial sensor against physical attacks because such protection would impair its functionality. Instead, it restricts the intended purpose to trusted environments without unauthorised physical access and informs users clearly about this restriction. In Example 67, a professional application stores sensitive data locally; rather than developing its own encryption, the manufacturer relies on the mature and regularly maintained encryption and key management functions of the operating system. Our reading: both examples show risk-based decisions, not carte blanche. In the first case, it is the more precisely defined intended purpose that carries the weight, not the notice alone.
Secure design and external dependencies
The first essential cybersecurity requirement requires products to be designed, developed and produced in such a way that they ensure an appropriate level of cybersecurity based on the risks; in practice, this is often called “security by design”. The guidance reads it as a catch-all requirement for all identified risks, in particular those not covered by the other essential requirements. Where all relevant risks are addressed through these other requirements, it is deemed fulfilled, which the Commission expects to be the case in most instances. Where additional risks remain, you need additional measures on the product.
The risk assessment also covers risks that originate outside the product, for example in external networks, third-party infrastructure or other systems. The Regulation does not require you to govern this environment, but it does require you to mitigate its risks through the design of the product. The guidance gives examples: against back-end systems that are not part of the product and could be misused to send malicious commands, cryptographic authentication of remote commands, integrity checks of configuration changes and security-relevant logs or alerts in the event of abnormal behaviour can help. If an external service fails, for example due to power or infrastructure outages, the product must not enter an insecure state. The Regulation does not regulate how the back-end infrastructure is organised, staffed or operated.
Due diligence for third-party components
- 1Derive needs from the risk assessment
Determine what the product requires from a component to meet its cybersecurity objectives, such as cryptographic functions, update mechanisms or secure communications. According to the guidance, this is the only way to fulfil the due diligence obligation.
- 2Verify in a risk-based manner
Check whether the component meets these needs. As evidence, the guidance cites documentation from the component manufacturer such as technical specifications, security documentation, or conformity documentation and other assurance documentation; where appropriate, your own tests supplement this.
- 3Treat as an externally supplied component
Components sourced from third parties and physically or logically integrated belong in the risk assessment. For compliance purposes, you treat them as externally supplied components whose properties you verify upon integration, because you have generally not designed them yourself.
- 4Feed the outcome into the demonstration of conformity
Due diligence is a separate legal obligation, but it underpins the demonstration that the product as a whole meets the essential requirements. Your own functionalities implement the requirements directly; third-party components are integrated with due diligence.
This expressly also applies to free and open-source components: in its open-source scenarios, the guidance repeatedly states that manufacturers integrating such components must fulfil their due diligence obligation, even if the component itself has not been placed on the market. The guidance does not use the term software bill of materials (SBOM); the Regulation itself requires, as part of vulnerability handling, that components be identified and documented, including by means of a software bill of materials. Our recommendation: bring the bill of materials, component requirements and due diligence evidence together in a single register so that procurement, development and product security work with the same data.
Product families: assess once, reflect the differences
Variants, models or configurations of the same product family do not each have to be treated as an entirely separate product if they share the same architecture, the same security-relevant design and the same intended purpose and are exposed to the same cybersecurity risks. Provided that all variants are adequately covered, the guidance considers a single risk assessment, a single set of technical documentation and a single conformity assessment procedure sufficient. A single EU declaration of conformity is also possible if it clearly identifies the variants covered. You remain responsible, however, for ensuring that the documentation accurately reflects the products actually placed on the market.
| Difference between variants | Security-relevant? | Consequence |
|---|---|---|
| Colour, form factor, memory size | No | No separate risk or conformity assessment required |
| Communication interfaces, software stacks | Yes | Reflect in the risk assessment and, where necessary, also in the conformity assessment and technical documentation |
| Update mechanisms, remote connectivity | Yes | Likewise; a single conformity assessment can only be relied on to the extent that the cybersecurity properties do not differ |
| New variant with new risks or a changed implementation of the requirements | Yes | Update the existing risk assessment and conformity documentation |
Remote data processing: the test and five use cases
Under the Regulation, a product with digital elements also includes its remote data processing solutions. The guidance translates the definition into a test with two cumulative questions and works through it using five fictional use cases. Who operates the back-end is irrelevant here; who designed and developed it is not.
The Regulation defines a product with digital elements as a software or hardware product including its remote data processing solutions. The risk assessment, the essential cybersecurity requirements and lifecycle obligations such as reporting actively exploited vulnerabilities and severe incidents therefore extend to the associated remote processing. Only the software elements are meant; the hardware on which the processing runs is not part of the product. Nor is the manufacturer’s entire IT meant, but only the software that enables the execution of the solution in question.
The test
- 1Does the processing take place at a distance?
Typically outside the user environment or an organisation’s operational environment, wired or wireless, in the cloud or close to the device at the edge. According to the guidance, this criterion is relevant but not sufficient and requires a case-by-case assessment by the manufacturer.
- 2Would the product lose one of its functions without it?
This means any function, not only the core functionality or the intended purpose, and thus also functions that support the product’s overall performance. The guidance cites sending commands to a device, synchronising files, onboarding users, configuration, automated distribution of updates and identity and access management.
- 3Was the software designed and developed by the manufacturer or under its responsibility?
Yes for in-house development and for development by a service provider based on your designs and specifications. No for merely licensing an existing product or service that a provider offers to all its customers, or slightly modified versions thereof.
- 4Outcome
Both questions answered in the affirmative: a remote data processing solution and thus part of the product. Function question answered in the negative: not a remote data processing solution, but the risks of the processing belong in the risk assessment. Function question affirmative, development question negative: a third-party solution that you treat like a component, with an assessment of the integration risks, measures on the product and due diligence.
Two clarifications help with borderline cases. If a function can be triggered both remotely and manually — the guidance cites a light bulb switched via an app or a switch — the remote processing can nevertheless be a remote data processing solution. Websites only count if they enable or support a product function, such as an authentication portal that issues credentials or tokens required for operation. Product information or a redirect from the product is not sufficient.
Development counts, operation does not
According to the guidance, who operates the solution is not decisive for the classification. If a third party operates a solution that you designed and developed, you remain responsible. Nor do remote data processing solutions have to run in a third-party cloud: servers on your premises, on-premises and private cloud solutions can fall within the definition just as much as the public cloud. The guidance distinguishes between cloud models: your own software on a third party’s infrastructure (IaaS) and your own application, whether created in-house or acquired, on a third party’s platform (PaaS) can be remote data processing solutions. A fully developed third-party application (SaaS) that you can configure only to a limited extent is not. Where the hypervisor, the platform operating system or the SaaS application is integrated into the product in a security-relevant manner, you treat it like a third-party component, with diligence proportionate to the risk. The provider’s security functions under the shared responsibility model can help here.
What is not remote data processing
- The manufacturer’s internal systems such as human resources, payroll and customer relationship management
- Continuous integration and continuous delivery (CI/CD) pipelines and the distribution of security updates to edge locations
- Auditing and testing systems for penetration testing, threat hunting and red teaming — even though effective and regular tests are themselves a requirement of the Regulation
- Remote analysis of telemetry data collected purely for statistical purposes or future product development
- Websites containing only product information or instructions, even if the product refers to them
- Downstream back-end systems with which the product does not interact directly
- Networks that serve only as a communication channel, such as cellular networks, ethernet cables, routers or Wi-Fi
Outside the definition does not mean outside the risk assessment. Even if remote processing does not support a product function, you must check whether it poses risks to the product and, where necessary, mitigate them at product level. Downstream back-end systems remain external dependencies. One borderline case deserves attention: the guidance cites the automated distribution of updates as a possible product function, but the distribution of security updates to edge locations as an internal system. Our reading: what matters is whether the product itself interacts with the service to perform one of its functions.
Five use cases from the guidance
| Case | Would a function be lost without the processing? | Designed and developed by the manufacturer? | Outcome and consequence |
|---|---|---|---|
| Mobile banking application: banking interface | Yes: authentication, submission of instructions, return of status | Yes, under the responsibility of the financial entity | Remote data processing solution; include in the risk assessment and the essential requirements for the product as a whole |
| Mobile banking application: account management and ledger | Availability needed, but no direct interaction with the app | Not relevant | Not a remote data processing solution, but an external dependency: strong authentication of back-end interfaces, integrity protection of transaction data, secure channels, verification of responses |
| Mobile banking application: support chat as SaaS | Yes | No: code, interface and operation lie with the third-party provider | Not a remote data processing solution, treated like a third-party component: isolate the chat from core functions, control data flows, validate content, exercise due diligence |
| Smart thermostat | Yes: exchanging commands, storing preferences | Yes, operated on third-party IaaS | Remote data processing solution; document the solution and the IaaS dependency, including contract details, for example by requesting evidence that NIS 2 obligations have been met |
| E-reader | Yes: access to purchased e-books | No: SaaS storage service that the provider makes available to all customers | Not a remote data processing solution, treated like a component: secure authentication, encryption, integrity protection of communications, due diligence in selection and integration |
| Industrial robot | Yes: without the calculated position, no picking up of parts | Yes, own remote service on third-party IaaS | Remote data processing solution; documentation including contract details, due diligence in selecting the IaaS provider |
| Cellular network | No: the network is only a communication channel | Not relevant | Neither a remote data processing solution nor a third-party component; no due diligence obligation towards the network operator |
The use cases in Section 8.3 of the guidance are fictional and archetypal. They illustrate the classification and do not replace a case-by-case assessment.
What to do for remote data processing solutions
- Indicate in the technical documentation that the product has remote data processing solutions or relies on third-party remote solutions, and describe them. If a solution serves several products, it belongs in each product’s documentation; its description may be reused.
- Draw the boundary precisely: the solution comprises the software modules responsible for the functionality and their interfaces with external services.
- Extend the risk assessment to three sources: your own solution, the reliance on third-party remote solutions and the product environment, including the underlying hardware.
- Use evidence that providers already have; the guidance lists, non-exhaustively, evidence relating to Implementing Regulation (EU) 2024/2690 and to DORA, statements of conformity or certificates under a European certification scheme, and ISO/IEC 27017:2015 and ISO/IEC 27001:2022.
- Embed security guarantees in service level agreements, including an assurance that the provider handles vulnerabilities adequately, and ensure that providers inform you about changes.
- Assess changes at the provider: a major change to its solution is not a substantial modification of your product, but may require a revised risk assessment, adjusted controls or a change of provider.
Reporting obligations and vulnerability handling: mandatory since 11 September 2026
Since 11 September 2026, manufacturers have had to report actively exploited vulnerabilities and severe incidents — including for products that have long been on the market. The guidance clarifies when the deadline starts to run, which vulnerabilities originating from third-party components must be reported and what counts as a known exploitable vulnerability before placing on the market.
For manufacturers, the reporting obligations are the part of the Regulation that already applies. As the guidance confirms, since 11 September 2026 they have covered all products within scope, including those placed on the market before 11 December 2027. And unlike the vulnerability handling obligations, they continue to apply after the support period has ended. Conversely: for products placed on the market before 11 December 2027 or whose support period has ended, you do not have to comply with the vulnerability handling requirements — but you do have to comply with the reporting obligations.
24 hEarly warning notification with limited information, counted from awareness
72 hNotification with additional information, counted from awareness
14 daysComplete report on the actively exploited vulnerability, after a corrective or mitigating measure is available
1 monthComplete report on the severe incident, counted from the 72-hour notification
What is reported and to whom
You must report every actively exploited vulnerability contained in your product and every severe incident having an impact on the security of the product as soon as you become aware of it. The recipients are, simultaneously, the CSIRT designated as coordinator and ENISA. Under the Regulation, these notifications are submitted via ENISA’s single reporting platform; the coordinating CSIRT in Germany is CERT-Bund at the BSI. The notifications must be updated progressively as your investigation advances. Open-source software stewards must also report actively exploited vulnerabilities, to the extent that they are involved in the development of a product. According to the guidance, this applies once active exploitation has been established, not merely because a vulnerability exists in the codebase.
When the deadline starts
- 1Suspicion or report
You detect a suspicious event yourself, or a third party brings it to your attention: an individual, a customer, an entity, an authority, a media organisation or another source.
- 2Immediate initial assessment
Assess the report immediately. The guidance places the emphasis on prompt action, particularly where the vulnerability may pose a significant risk.
- 3Awareness with a reasonable degree of certainty
You are aware as soon as, after the initial assessment, you establish with a reasonable degree of certainty that a vulnerability in the product is being actively exploited or that a severe incident has compromised the security of the product. The deadlines run from this point.
- 4Early warning, notification, complete report
Early warning notification without undue delay, at the latest 24 hours after becoming aware; notification without undue delay, at the latest 72 hours after becoming aware; complete report at the latest 14 days after a corrective or mitigating measure is available, or, for severe incidents, at the latest one month after the 72-hour notification. The Regulation calls the complete report the “final report”.
- 5Inform users
Impacted users and, where appropriate, all users must be informed, in a risk-based and proportionate manner. This does not mean publishing details indiscriminately: particularly in sensitive or essential environments, you may limit detailed information to the customers concerned. If you do not inform users in a timely manner, the CSIRTs that received the notification may inform the users themselves.
Two clarifications spare you unnecessary notifications. First, there is no retroactive reporting: if you were already aware of the active exploitation before 11 September 2026, you do not have to report it. If at that time you knew only of the vulnerability but not of its exploitation, the reporting obligation applies as soon as exploitation subsequently occurs or you become aware of it. Second, you report only actively exploited vulnerabilities contained in your own product, even if they originate from a third-party component. If a component’s vulnerability cannot be exploited in your product, for example because the vulnerable code is not reachable, or if it has not been exploited there, you have no reporting obligation. Voluntary notification remains possible. Handling the vulnerability and reporting it to the component manufacturer are governed by the other manufacturer obligations.
Vulnerabilities in components: report upstream, share fixes
The Regulation requires manufacturers to report vulnerabilities in integrated components to the person or entity manufacturing or maintaining the component, and to share security fixes they have developed themselves with that person or entity. The guidance refers to this as reporting upstream and sharing upstream. It limits reporting upstream to the version of the component that you integrate and sets out clear exceptions for both obligations; it does not specify a deadline.
| Situation | Obligation according to the guidance |
|---|---|
| Vulnerability in the integrated version of the component | Report upstream via the maintainer’s security policies, coordinated vulnerability disclosure processes or designated reporting channels |
| The maintainer is demonstrably already aware of the vulnerability | No reporting obligation; it is recommended to avoid duplicate reports by checking beforehand, for example in public vulnerability databases, the project’s security advisories or issue trackers |
| The vulnerability arises only from the integration with your own code or other components | No reporting obligation; if the integration reveals previously unknown characteristics of the component, communicating them is recommended |
| The component no longer has a maintainer, or you no longer rely on the original maintainer | No reporting obligation; it is recommended to inform users of the component via mailing lists, tickets or public vulnerability records |
| Your own security fix for the component | Share it with the maintainer, where appropriate in a machine-readable and easily integrable form, and for free and open-source software in a licence-compatible manner; you do not have to ensure that it is accepted |
| The maintainer has a fix, you mitigate differently, for example via configuration | No obligation to share and no obligation to adopt the maintainer’s fix; sharing is recommended where it helps the maintainer |
Known exploitable vulnerabilities before placing on the market
The essential cybersecurity requirements include making products available on the market on the basis of the risk assessment and, where applicable, without known exploitable vulnerabilities. A vulnerability is exploitable if an adversary can effectively use it under practical operational conditions; what works only in a lab or in a simulation does not count. According to the guidance, it is known when it is listed in relevant publicly accessible vulnerability databases, such as the European vulnerability database. It may also be known from non-public sources — from coordinated disclosure by security researchers or from your own testing and analysis, including with AI-powered services — or from prominent reporting in reliable specialist or mass media. Check promptly whether a finding is accurate and affects your product. The obligation is risk-based: if a vulnerability becomes known shortly before the product enters the distribution chain, you decide on the basis of the risk assessment whether the product can be securely placed on the market or whether you fix it first. The criteria are severity, exploitability, potential impact and the risks once in use; you may also take into account the risk of postponement, for example where the new version fixes other exploitable vulnerabilities or is needed for the continued operation of critical systems. This is not a blanket exemption; after placing on the market, the obligation to handle vulnerabilities effectively remains.
Testing without a rigid schedule
Throughout the entire support period, the Regulation requires effective and regular tests and reviews. According to the guidance, regular does not mean mechanically repeating an unchanged test campaign at fixed intervals. It means regularly reviewing whether new input, such as newly identified threats or newly discovered vulnerabilities, requires adapted tests, and then executing them, including as regression tests where the product has changed. Frequency, depth and content depend on the risk profile, the evolution of the product and the threat landscape. If the review identifies nothing new, no additional tests are needed; the other vulnerability handling obligations continue to apply in parallel.
Interplay with other legislation: vehicles, type-examination certificates and what is still to come
The guidance interprets only the Cyber Resilience Act, but it expressly addresses two interfaces with other product legislation: vehicle components and existing EU type-examination certificates. For both, it draws narrow boundaries. The Commission leaves the door open to further guidance, for example on the relationship with the AI Act and with DORA.
At the outset, the Commission clarifies that its guidance concerns only the Cyber Resilience Act and does not apply to other EU legislation. Section 9.3 of the guidance addresses two interfaces in detail: the exemption for vehicles and their components, and the continued validity of examination certificates issued under other Union harmonisation legislation. Further references to NIS 2, DORA, the GDPR and the Cybersecurity Act are scattered throughout the text.
Vehicles and their components
The CRA does not apply to products to which Regulation (EU) 2019/2144 on the type-approval of motor vehicles of categories M, N and O applies. Products within the scope of Regulation (EU) No 168/2013 for L-category vehicles have been excluded by the Commission through Delegated Regulation (EU) 2025/1535. The vehicles themselves are therefore not subject to the Cyber Resilience Act. For systems, components, separate technical units, parts and equipment, which the guidance refers to collectively as components, it draws a narrow boundary: these are exempt only if they are exclusively designed and constructed for integration into such vehicles. Whether you supply directly to the vehicle manufacturer or to another economic operator in the automotive supply chain makes no difference, as long as the component is clearly intended and suitable only for ultimate integration into these vehicles.
| Scenario | Classification according to the guidance |
|---|---|
| Component exclusively designed and constructed for integration into such vehicles, distributed via restricted B2B channels of the automotive supply chain | Not covered by the Cyber Resilience Act; the restricted channels indicate exclusivity |
| The same component, supplied to a supplier instead of directly to the vehicle manufacturer | Likewise not covered, as long as it is clearly intended and suitable only for ultimate integration into such vehicles |
| Generic component that can be integrated into different types of products | Covered |
| Component that is not exclusively suitable, offered through general retail outlets or online channels for the general public | Covered, irrespective of any statements concerning its intended use |
What matters are the objective conditions under which the component is made available, not the labels. A notice saying “for vehicle installation only” does not help if the component is technically suitable elsewhere too and is sold freely. Our recommendation for suppliers: document the scope decision for each component, including its suitability and distribution channels, and review mixed portfolios component by component.
Existing type-examination certificates: a bridge until 11 June 2028
Under the Regulation’s transitional provision, EU type-examination certificates and approval decisions issued regarding cybersecurity requirements under other Union harmonisation legislation remain valid until 11 June 2028, unless they expire earlier or the other legislation specifies otherwise. The guidance interprets this bridge narrowly. It holds only for the cybersecurity risks and requirements of the legislation under which the certificate was issued. For those risks, you do not have to reassess conformity solely for the purposes of the Cyber Resilience Act for products that you place on the market from 11 December 2027 until 11 June 2028. This applies to mandatory procedures involving a notified body as well as to voluntarily chosen ones. Beyond 11 June 2028, you cannot rely on it for the purposes of the Cyber Resilience Act, even if the certificate remains valid for longer under other legislation.
| Basis of the certificate | Covered, to the extent addressed in the certificate | Possible gaps according to the guidance |
|---|---|---|
| Radio Equipment Directive Delegated Act, Delegated Regulation (EU) 2022/30 | Network protection, protection of personal data and privacy, prevention of fraud | Vulnerability handling processes, data minimisation, reduction of the attack surface and other product-specific aspects |
| Machinery Regulation (EU) 2023/1230, fully applicable as of 20 January 2027 | “Protection against corruption”, i.e. against tampering, as well as safety and reliability of control systems under the relevant sections of the Machinery Regulation | All other cybersecurity risks not covered by the certificate |
For both cases, the guidance stresses the same point: a valid certificate does not replace the comprehensive cybersecurity risk assessment under the Cyber Resilience Act and does not exempt you from its other obligations. A certificate under the Radio Equipment Directive does not in itself demonstrate full compliance with the Cyber Resilience Act. It serves as evidence for the risks it covers; you must assess and mitigate gaps, irrespective of whether the certificate remains valid. Products placed on the market before 11 December 2027 are subject to the legislation applicable at that time, unless they are substantially modified on or after that date.
Other legal acts referred to in the guidance
- NIS 2 Directive and Implementing Regulation (EU) 2024/2690: cloud computing services fall under NIS 2. Manufacturers can use evidence that cloud providers fulfil these obligations as due diligence evidence, and the interpretation of “becoming aware” in the reporting obligations is aligned with the Implementing Regulation.
- GDPR: for the start of the reporting deadlines, the guidance is also aligned with Guidelines 9/2022 on personal data breach notification, so that manufacturers with parallel reporting obligations can work consistently.
- DORA: evidence of fulfilment of DORA obligations is among the assurance artefacts that can be reused for third-party remote services.
- Cybersecurity Act, Regulation (EU) 2019/881: the basis for European cybersecurity certification schemes, which play a role in conformity assessment and as evidence.
- Implementing Regulation (EU) 2025/2392: technical description of the categories of important and critical products.
- The Commission’s “Blue Guide”: reference point for placing on the market, making available on the market and substantial modifications, and for the manufacturer’s responsibility for all risks, even when applying harmonised standards.
Outlook: AI Act and DORA
On the basis of the Regulation, the Commission may consider issuing further guidance, including for manufacturers that are also subject to other Union harmonisation legislation or related Union legal acts. As possible topics, it cites the interplay with the AI Act, Regulation (EU) 2024/1689, and with DORA, Regulation (EU) 2022/2554 on digital operational resilience for the financial sector. It does not give a timetable. Artificial intelligence appears only marginally in the current guidance: your own testing with AI-powered services can make a vulnerability “known”. Our assessment: anyone integrating AI functions into products with digital elements or supplying financial entities should not wait for this guidance, but should set up risk assessment, reporting channels and evidence now in such a way that they serve several regulatory frameworks.
What to do now: our roadmap to 11 December 2027
This chapter is a recommendation from VamiSec, not a requirement of the Commission. It translates the guidance into a sequence based on the dates of the Regulation: first what already applies, then the decisions on which everything else depends.
| Date | What applies |
|---|---|
| 11 June 2026 | The provisions on the notification of conformity assessment bodies apply |
| 11 September 2026 | Reporting obligations for actively exploited vulnerabilities and severe incidents apply, including for products already on the market |
| 17 November 2026 | Commission guidance published in the Official Journal (C/2026/6013) |
| 20 January 2027 | Machinery Regulation fully applicable, relevant for manufacturers of machinery |
| 11 December 2027 | Full application: manufacturer obligations, conformity assessment, EU declaration of conformity and CE marking |
| 11 June 2028 | Existing EU type-examination certificates under other product legislation can be relied on for the purposes of the Cyber Resilience Act until this date at the latest |
Immediately: make reporting operational — mandatory since 11 September 2026
- Define and publish intake channels for reports, for customers, security researchers, authorities and partners.
- Introduce triage with a documented initial assessment, using the criterion “reasonable degree of certainty” and a timestamp for the moment of awareness.
- Create a reporting runbook with roles, deputies, deadline tracking and templates for the early warning notification, the notification and the complete report.
- Prepare the reporting route via the single reporting platform provided for by the Regulation, and identify the competent CSIRT designated as coordinator.
- Include all products within scope, including products already on the market and products no longer supported.
- Triage vulnerabilities in third-party components: is the vulnerable code reachable in your product, and is it being actively exploited there?
- Define criteria for informing users: recipients, level of detail, timing.
- If you act as an open-source software steward: set up a publicly accessible channel for reports of active exploitation, because stewards usually learn of it through third parties.
Within 90 days: product inventory and scope decisions
- Create a product inventory covering hardware, software, variants, versions, separately offered modules, companion apps and back-ends.
- Delineate software: locally executed software and apps are products with digital elements. Web applications accessed only through a browser generally are not, but may be covered as remote data processing if they support a product function.
- Assign companion software: software necessary to operate, configure, control or use a device is part of the product, even if it is obtained through an app store.
- Check the data connection: does the product exchange digitally encoded information, or does it merely trigger a function via electrical signals?
- Clarify the open-source role for each project: manufacturer, open-source software steward or mere contributor.
- Record all remote processing and document the two remote data processing test questions for each case.
- Flag legacy designs: products designed before 11 December 2027 but placed on the market after that date.
Within six months: risk assessment, classification, conformity route
- Set up the risk assessment for each product or product family, with residual risk criteria based on the product rather than on risk appetite or costs.
- Determine the core functionality of each product and each separately offered module and compare it against the category descriptions in Implementing Regulation (EU) 2025/2392.
- Determine the conformity route, for class I depending on whether references of harmonised standards have been published in the Official Journal; where a third party is involved, plan early with a conformity assessment body.
- Operationalise due diligence for components: requirements per component, evidence from the component manufacturer, risk-based checks and the software bill of materials required by the Regulation.
- Treat cloud and SaaS providers that are not your own remote data processing solution like components: obtain evidence and lay down security guarantees and change notifications in contracts.
- Compare existing certificates under radio equipment or machinery legislation against the risk assessment and plan for the gaps.
- Include remote data processing solutions and their interfaces in the risk assessment, security requirements, testing and reporting process.
Within twelve months: processes, support period, documentation
- Add an update classification to the release process: does an update change the intended purpose, or does it introduce new or increased risks not covered by the risk assessment? Scale and complexity are not the yardstick.
- Justify the support period for each product on the basis of the expected use time, make the end date visible with month and year at the time of purchase and implement the end-of-support notification.
- Define a version strategy: if you only want to patch the latest version, you need a free upgrade path without additional costs for users.
- Introduce a release check for known exploitable vulnerabilities with a documented, risk-based decision.
- Describe reporting upstream and sharing security fixes with component manufacturers and maintainers as a procedure, with a directory of channels, duplicate checks and licence checks.
- Switch the test concept to triggers: new threats, newly discovered vulnerabilities, product changes.
- Build up the technical documentation for each product: core functionality, risk assessment, components, remote data processing, support period, applied standards and additional measures.
- Collect evidence: test reports, due diligence records, release records and reporting drills.
By 11 December 2027: conformity assessment, EU declaration of conformity, CE marking
From 11 December 2027, the Regulation requires that, before a product is placed on the market, the applicable conformity assessment procedure has been carried out, the EU declaration of conformity drawn up and the CE marking affixed. According to the guidance, this also applies to legacy designs that continue to be placed on the market without redesign. For hardware, each individual unit counts; a standalone software version, by contrast, is considered placed on the market when it is first offered. Our recommendation: plan a buffer so that the assessment is completed before the first unit is placed on the market after the cut-off date. For product families, you can use a single risk assessment, a single set of technical documentation and a single EU declaration of conformity if the variants are clearly identified in them and do not differ in their cybersecurity properties.
The guidance is not binding; an authoritative interpretation of the Regulation can only be given by the Court of Justice of the European Union, and the examples in the guidance do not replace a case-by-case assessment. This analysis does not constitute legal advice.
Integrated gap analysis
CRA gap analysis: where do your products stand against the guidance?
In around 15 minutes through up to 45 questions in eight action areas — from scope through open source, risk assessment and support period to remote data processing and reporting. Four profile inputs control the questions and the target level, so that the assessment remains proportionate to your portfolio. Each question cites its reference in the guidance. Your result stays in your browser; levels, weighting and target levels are a classification by VamiSec.
0 / 45 questions answered
Your evaluation
Not all questions answered yet — the evaluation is provisional.
Your biggest gaps
Not all questions answered yet — the evaluation is provisional.
The button takes you to the whitepaper form, where you can optionally send your result along.
Your answers stay in your browser; nothing is stored or transmitted. Only when you click “Discuss your result with VamiSec” is the summary passed to the form — and it is sent only if you choose to do so there. The result link contains your profile and levels as codes; share it only selectively.
Free whitepaper
CRA Guidance for CISOs — the Commission’s interpretation in practice
The whitepaper translates the guidance published on 17 November 2026 into concrete review steps for CISOs and product owners: from scope through open source, substantial modification and support period to core functionality, remote data processing and reporting obligations — with references, a gap analysis checklist and a roadmap to 11 December 2027. Available in German and, as “CRA Guidance for CISOs”, in English.

18 chaptersGerman and EnglishAs of 21 November 2026PDF, free of charge
- Scope and open source: software, web applications, devices with a companion app, legacy designs, maintainers, contributors and stewards
- Substantial modification, spare parts, support period and core functionality — with the examples from the guidance
- Remote data processing with all five use cases and the reporting obligations that have applied since 11 September 2026
- Gap analysis checklist and roadmap to 11 December 2027, separated into Regulation, Commission interpretation and VamiSec recommendation
What the whitepaper “CRA Guidance for CISOs” contains
Scope and open source
When software, web applications, source code, devices with a companion app and legacy designs fall under the CRA — and when free and open-source software is covered: maintainers, contributors, monetisation and the role of the open-source software steward, each with the relevant examples from the guidance.
Substantial modification, support period, core functionality
When a software update or a spare part leads to a substantial modification, how to derive the support period from the expected use time and how the core functionality determines the class and the conformity assessment procedure.
Remote data processing and reporting obligations
The two-condition test for remote data processing with all five use cases from the guidance, treating cloud providers like third-party components and the reporting obligations that have applied since 11 September 2026 — with the start of the deadline, third-party components and reporting upstream.
Gap analysis checklist and roadmap
A self-assessment checklist along the eight topic areas of this page and a roadmap to the date of application on 11 December 2027 — clearly separated into what the Regulation requires, how the Commission interprets it and what VamiSec recommends.
Foundation
The Regulation first, then the interpretation: the CRA deep dive
The guidance presupposes the Regulation: it explains selected key provisions and expressly does not cover the entire scope. The basics — product classes, essential cybersecurity requirements, conformity assessment and CE marking, reporting deadlines and obligations across the life cycle — are explained in our deep dive on the EU Cyber Resilience Act. You can find the official text of the guidance in all official languages in the Official Journal on EUR-Lex.
Regulation (EU) 2024/2847binding obligations, classes and deadlines
C/2026/6013the Commission’s interpretation, not binding
Implementing Regulation (EU) 2025/2392technical description of the product categories
- The Regulation has been in force since 10 December 2024; the reporting obligations have applied since 11 September 2026, and the other obligations of economic operators apply from 11 December 2027.
- The Regulation requires the Commission to cover at least four topics: scope including remote data processing and open source, support periods, substantial modification and the interplay with other EU legislation.
- As early as 3 December 2025, the Commission had published frequently asked questions to help with preparation; the guidance followed after a public consultation from 3 March to 13 April 2026.
- According to its own statement, the Commission may consider further guidance, for example on the interplay with the AI Act and DORA.
The text published in the Official Journal is authoritative. The German language version differs from the English one in some places — for example where it speaks of “Inkrafttreten” (entry into force) when the date of application on 11 December 2027 is meant.
VamiGRC
How to implement this with VamiGRC
VamiGRC takes your products through defined release gates, attaches CRA requirements to the right phase and links SBOM, suppliers and evidence – all the way to a prepared notification for BSI and ENISA.
Glossary
Glossary: the terms of the CRA guidance
36 terms, mostly in the official English terminology of the guidance, explained in our own words — with the German term for comparison with the German language version.
36 of 36 terms
- Cyber Resilience Act (CRA)Cyberresilienzverordnung (CRA)
- Regulation (EU) 2024/2847 on horizontal cybersecurity requirements for products with digital elements. It has been in force since 10 December 2024; its reporting obligations have applied since 11 September 2026, and the other obligations of economic operators apply from 11 December 2027.
- Commission guidanceLeitlinien der Kommission
- Communication C/2026/6013, in which the Commission sets out its interpretation of key provisions of the Cyber Resilience Act, published in the Official Journal on 17 November 2026. It is not binding; only the Court of Justice of the EU can give an authoritative interpretation.
- Date of applicationGeltungsbeginn
- The date from which the obligations of the Regulation apply in full: 11 December 2027. In several places, the German language version of the guidance speaks of “Inkrafttreten” (entry into force) but means this date — the Regulation has already been in force since 10 December 2024.
- Product with digital elementsProdukt mit digitalen Elementen
- A software or hardware product, including its remote data processing solutions and components placed on the market separately. It is covered if its intended purpose or reasonably foreseeable use includes a data connection to a device or network.
- Placing on the marketInverkehrbringen
- The first making available of a product on the Union market; every subsequent supply of the same product is a making available on the market. For standalone software, all copies of a version are considered placed on the market with the first offer.
- Commercial activityGeschäftstätigkeit
- Supply of a product in a business-related context, whether in return for payment or free of charge. Whether it exists must be assessed case by case, for example by regularity, the characteristics of the product and the intention of the supplier; for free and open-source software, monetisation is the main deciding factor.
- ManufacturerHersteller
- A person who supplies a product with digital elements under its own name or trademark in the course of a commercial activity and thereby places it on the market. A person who substantially modifies a product and makes it available on the market is also considered a manufacturer.
- Importers and distributorsEinführer und Händler
- Importers place products from manufacturers outside the EU on the Union market, distributors make them available in the supply chain; both have their own obligations. If they carry out a substantial modification, they are themselves considered manufacturers.
- Standalone softwareEigenständige Software
- Software that is not combined with hardware into one product, such as an app or a desktop program — distributed digitally or on a physical medium. The rules on placing on the market apply to it per version; software belonging to a device follows the timing of the hardware.
- Data connectionDatenverbindung
- Transmission of information in binary form that a source deliberately encodes according to a defined scheme and that a recipient can read as data. In the Commission’s view, signals that only trigger or power a function are not a data connection.
- Complex systemKomplexes System
- An interplay of several hardware and software elements that is placed on the market as a single product, often with long development and operating periods. It is not exempt; technical constraints must be documented and mitigated by compensatory measures.
- Remote data processingDatenfernverarbeitung
- Data processing at a distance for which the manufacturer has designed and developed the software itself or under its responsibility, and without which the product could not perform one of its functions. It is part of the product — regardless of who operates it.
- Free and open-source software (FOSS)Freie und quelloffene Software
- Software whose licence freely permits use, modification and redistribution and whose source code is openly shared. Source code that is accessible only to paying customers or a limited group does not meet the definition.
- MaintainerMaintainer
- A person who publishes free and open-source software and exercises primary control over its development, releases and distribution; the software is under that person’s responsibility. In some places, the German language version of the guidance calls maintainers “Wartungsbetriebe”.
- ContributorContributor
- A person who contributes source code to free and open-source software without controlling releases, roadmap or governance. To that extent, contributors are not subject to the Regulation — not even with commit access.
- Open-source software stewardVerwalter quelloffener Software
- A legal person that systematically supports, on a sustained basis, the development of free and open-source software that is intended for commercial activities but not placed on the market — a foundation, for example. A separate, lighter regime of obligations applies to it; the reporting obligations depend on the type of support.
- Due diligenceSorgfaltspflicht
- The manufacturer’s obligation, when integrating third-party components — including free and open-source software — to determine what they must deliver for the security of the product and to verify this on a risk basis. This does not make the manufacturer responsible for the conformity of the component itself.
- Substantial modificationWesentliche Änderung
- A change after placing on the market that affects compliance with the essential cybersecurity requirements or modifies the intended purpose for which the product was assessed. For software updates, what counts is whether new or increased risks arise that the risk assessment does not cover; the consequence is a new placing on the market.
- Security updateSicherheitsaktualisierung
- An update whose primary purpose is to reduce cybersecurity risks. As a rule, it is not a substantial modification — unless it changes the intended purpose or introduces new risks, for example through new data flows or externally reachable interfaces.
- Spare partErsatzteil
- A part that replaces an identical component, is manufactured according to the same specifications and is supplied specifically to repair or extend the durability of a product already placed on the market; it does not fall under the Regulation. The yardstick for “identical” is the functional role and the security-relevant characteristics.
- Support periodUnterstützungszeitraum
- The period during which the manufacturer effectively handles vulnerabilities in a product, including its components. It reflects the expected use time and is, as a rule, at least five years — a minimum, not a standard.
- Core functionalityKernfunktion
- The main features and technical capabilities without which a product could not fulfil its intended purpose. Every product has exactly one core functionality; it determines whether the product is important, critical or part of the default category.
- Important products with digital elementsWichtige Produkte
- Products whose core functionality corresponds to a category on the list of important products, divided into class I and class II — for example operating systems, routers or firewalls. They are subject to stricter conformity assessment procedures than the default category.
- Critical products with digital elementsKritische Produkte
- Products whose core functionality corresponds to a category on the list of critical products. According to the guidance, their conformity assessment always requires the involvement of a third party.
- Default categoryStandardkategorie
- Not a term of the Regulation but the guidance’s usage for products without a core functionality from the lists of important or critical products. For them, self-assessment under the internal control procedure (module A) is always sufficient.
- Essential cybersecurity requirementsGrundlegende Cybersicherheitsanforderungen
- The Regulation’s requirements for the design, development and production of a product and for the handling of vulnerabilities. Which of them apply in the individual case and how they are implemented follows from the cybersecurity risk assessment.
- Cybersecurity risk assessmentBewertung der Cybersicherheitsrisiken
- A mandatory assessment in which the manufacturer identifies the relevant risks of its product and derives appropriate measures — taking into account the intended purpose, reasonably foreseeable use, operational environment, assets to be protected and expected use time. It must be kept up to date during the support period.
- Residual cybersecurity riskRestrisiko
- The risk that remains after the measures have been implemented. The yardstick is an appropriate level of cybersecurity — not internal risk appetite, business strategy or costs; it cannot be shifted to users by means of a warning notice.
- Family of products with digital elementsProduktfamilie
- Variants, models or configurations with the same architecture, the same security-relevant design, the same intended purpose and the same risks. One risk assessment, one set of technical documentation and one conformity assessment procedure may suffice for them; security-relevant differences such as different interfaces or update mechanisms must be reflected separately.
- Conformity assessment procedureKonformitätsbewertungsverfahren
- The procedure used before placing on the market to demonstrate that a product meets the essential requirements — from self-assessment under module A to procedures with third-party involvement under module B+C or module H. Which one applies depends on the core functionality.
- Presumption of conformityKonformitätsvermutung
- Anyone who applies a harmonised standard whose reference has been published in the Official Journal is presumed to conform with the requirements it covers. The presumption extends only as far as the standard; for additional functionalities that are not covered, compliance must be demonstrated in another way.
- Actively exploited vulnerabilityAktiv ausgenutzte Schwachstelle
- A vulnerability that is not merely exploitable but is actually being exploited. Manufacturers report it as soon as, after a prompt initial assessment, they know with a reasonable degree of certainty that a vulnerability in their product is being actively exploited — mandatory since 11 September 2026.
- Severe incidentSchwerwiegender Sicherheitsvorfall
- An incident that has an impact on the security of a product and is considered severe under the criteria of the Regulation. It must also be reported; the complete report follows no later than one month after the 72-hour notification.
- Known exploitable vulnerabilityBekannte ausnutzbare Schwachstelle
- A vulnerability that is exploitable under practical operational conditions and known — for example from public vulnerability databases, from coordinated disclosure, from the manufacturer’s own tests, including AI-assisted ones, or from prominent reports by reliable media. Whether the product can nevertheless be placed on the market is decided by the manufacturer on a risk basis.
- Reporting upstreamVorgelagerte Meldung
- The obligation to report a vulnerability in an integrated component to the person or entity manufacturing or maintaining the component, and to share one’s own fixes. Reporting is not required, for example, where that person or entity demonstrably already knows about the vulnerability or the component is no longer maintained.
- Single reporting platformEinheitliche Meldeplattform
- The platform established by ENISA through which notifications under the Regulation go to the CSIRT designated as coordinator — in Germany, CERT-Bund at the BSI — and to ENISA. The guidance itself does not name the platform, only the two recipients.
FAQ
Frequently asked questions on the Commission’s CRA guidance
Short answers to typical questions from product management, product security, engineering and compliance — based on the guidance of 17 November 2026, not legal advice.
No. The guidance is not binding on economic operators; only the Court of Justice of the European Union can give an authoritative interpretation of the Cyber Resilience Act. However, it reflects the Commission’s interpretation and is expressly also intended to support the work of market surveillance authorities, notifying authorities and notified bodies, so that the CRA is enforced uniformly across the Union. The examples do not replace a case-by-case assessment. Our recommendation: where you depart from the Commission’s line, document the reasons.
Not in itself. In the Commission’s view, software that is executed remotely and that users merely access — for example a web application or progressive web app used exclusively in the browser — is not a product with digital elements; nor are websites. However, it is covered as remote data processing if a product could not perform one of its functions without it and the manufacturer developed the software itself or under its responsibility. Apps, desktop clients and browser extensions for your service, by contrast, are separate products.
When it introduces new or increased cybersecurity risks that the risk assessment does not cover, or changes the intended purpose. The yardstick is the risk, not the scope or complexity: as non-exhaustive test criteria, the guidance names new threat vectors, new attack scenarios and a changed likelihood or impact of known attack scenarios. Even a small feature such as “remember me” can be substantial (Example 44 of the guidance). Security updates are, as a rule, not substantial — unless they change the intended purpose or introduce new risks that the risk assessment did not anticipate, for example through substantially changed data flows or new externally reachable interfaces.
As long as the product is expected to be in use — as a rule at least five years, shorter only if the expected use time is shorter. The Commission stresses that five years is a minimum and not a standard; criteria include reasonable user expectations, the nature and intended purpose of the product and relevant Union law on product lifetime. You state the end date at the time of purchase, at least with month and year. After a substantial modification, you reassess: if the factors determining the use time remain the same, the original period continues — even if less than five years remain.
Contributors who only contribute code and control neither releases nor roadmap or governance are not subject to the Regulation — not even with commit access. Responsibility lies with the maintainers. They become manufacturers if they monetise the software, for example through a price, a paid or enterprise version, the monetisation of other services or personal data as a condition of use; a donation link, optional support or sponsorship is not enough for this. Legal persons that provide sustained support for non-monetised software intended for commercial activities can be open-source software stewards with lighter obligations — and not-for-profit organisations that use all earnings after costs for not-for-profit objectives can be so even if they monetise the software.
Not necessarily. If you place units of a previously developed type on the market from 11 December 2027, the Regulation does not, in the Commission’s view, automatically require a redesign. What matters is an up-to-date cybersecurity risk assessment: if it demonstrates appropriate and effective measures, you may rely on it. You do not have to recreate historical documentation, and evidence can be grouped for product families. Conformity assessment, the EU declaration of conformity, CE marking and vulnerability handling nevertheless remain mandatory — and if the assessment reveals gaps, you need additional measures.
Only under certain conditions. Exempt are spare parts that replace identical components, are manufactured according to the same specifications and are supplied specifically to repair or extend the durability of a product already placed on the market — recognisable, for example, from the product reference in the order or offer, or from supply via after-sales service. “Identical” is measured by the functional role and security-relevant characteristics: a different chipset may be harmless, a different cryptography implementation not. Non-identical spare parts are separate products within the meaning of the Regulation.
When its core functionality corresponds to a category on the lists of important (class I or II) or critical products; the technical descriptions are set out in Implementing Regulation (EU) 2025/2392. According to the guidance, every product has exactly one core functionality, determined objectively rather than by marketing; an integrated operating system does not make a smartphone an operating system. For class I, self-assessment is sufficient only with a harmonised standard applied in full whose reference has been published in the Official Journal and that covers all risks of the core functionality. Class II and critical products require third-party involvement; important products placed on the market as free and open-source software, by contrast, may use the procedures of the default category.
Your backend is part of the product if two conditions are met together: without the processing, the product could not perform one of its functions, and the software was designed and developed by you or under your responsibility according to your specifications. Who operates it is irrelevant; your own servers or a private cloud also count. Typical examples are remote control, synchronisation, onboarding, configuration, update distribution and identity and access management. Not included are telemetry for statistics only, HR, CRM and CI/CD systems or standard SaaS that you buy in — you treat the latter like a third-party component.
Yes, since 11 September 2026 — for all products in scope, including those placed on the market before 11 December 2027 and also after the end of the support period. Actively exploited vulnerabilities and severe incidents are notified simultaneously to the CSIRT designated as coordinator and to ENISA: early warning within 24 hours, notification within 72 hours, complete report 14 days after a corrective or mitigating measure is available (vulnerability) or one month after the 72-hour notification (incident). The deadline runs from becoming aware with a reasonable degree of certainty after a prompt initial assessment. There is no retroactive reporting.
To the CSIRT and ENISA only if the vulnerability is being actively exploited in your product — but then also if it originates from a third-party component. If the vulnerable code is not reachable in your product or nothing has been exploited, there is no reporting obligation; you can still report voluntarily. Independently of this, the Regulation requires you to report vulnerabilities in integrated components to their manufacturer or maintainer and to share your own fixes. According to the guidance, reporting upstream is not required if the maintainer demonstrably knows about the vulnerability or the component is no longer maintained.
The final version is around 23 per cent longer: 257 instead of 232 paragraphs, 67 instead of 55 examples. New are a section distinguishing locally executed software from pure web applications, a section on the support period after substantial modifications and explanations on regular testing. The spare parts exemption is now linked to the supply context and security-relevant characteristics. Substantial modifications to existing products no longer automatically require full conformity of the entire product, as long as the product as a whole is not negatively affected. For the reporting obligations, the cut-off date, existing products and the exclusion of retroactive reporting were added.
Little: the guidance concerns the Cyber Resilience Act alone. The Commission keeps open the option of issuing further guidance, for example on the interplay with the AI Act and with DORA. DORA becomes concrete in one place: evidence that providers of remote services present on DORA or NIS 2 obligations can be reused by manufacturers when assessing third-party providers, as can evidence under ISO/IEC 27001 or 27017. By contrast, the chapter on other legislation expressly deals with vehicle components and with certificates under the Radio Equipment Directive and the Machinery Regulation.
The integrated gap analysis translates the guidance into review questions on eight topic areas — from scope to reporting. First you define your profile: product type, important or critical products, open source, legacy designs. Then you rate each question on a scale from 0 to 3 and receive a maturity level, gaps with measures and a roadmap for 30, 90 and 180 days. Your answers remain in your browser. The result is a self-assessment — not legal advice and not a conformity assessment.
Standards & Sources
The content on this page is based on the following publicly available guides and studies.
European Commission · 2026
Commission guidance on the application of Regulation (EU) 2024/2847 (Cyber Resilience Act), C/2026/6013 ↗
Primary source of this page — communication from the Commission, dated 9 November 2026, published in the Official Journal (C series) on 17 November 2026; 9 chapters, 257 paragraphs, 67 examples, 5 use cases; not binding
European Parliament and Council of the EU · 2024
Regulation (EU) 2024/2847 on horizontal cybersecurity requirements for products with digital elements (Cyber Resilience Act) ↗
Legal basis — OJ L of 20 November 2024, in force since 10 December 2024; notification of conformity assessment bodies from 11 June 2026, reporting obligations (Art. 14) since 11 September 2026, full application from 11 December 2027
European Commission · 2025
Commission Implementing Regulation (EU) 2025/2392 — technical description of the categories of important and critical products with digital elements ↗
Yardstick for mapping the core functionality; used in the guidance for operating systems, routers and antivirus software, among others (Examples 57, 58, 62, 63)
European Commission · 2025
Frequently asked questions (FAQs) on the implementation of the Cyber Resilience Act ↗
Published on 3 December 2025 and available via the Commission’s CRA pages; the guidance refers to them (para. 5)
European Commission · 2022
Commission Notice — The “Blue Guide” on the implementation of EU product rules 2022, 2022/C 247/01 ↗
OJ C 247 of 29 June 2022, p. 1; used by the guidance for placing on the market, commercial activity, substantial modifications and harmonised standards
European Parliament and Council of the EU · 2019
Regulation (EU) 2019/1020 on market surveillance and compliance of products ↗
Market surveillance framework to which products within the scope of the CRA are also subject (para. 3)
European Parliament and Council of the EU · 2022
Directive (EU) 2022/2555 on measures for a high common level of cybersecurity across the Union (NIS 2 Directive) ↗
In the guidance, the reference point for cloud computing services, the European vulnerability database and evidence from IaaS providers (paras 183, 233; use cases 8.3.2, 8.3.4)
European Commission · 2024
Commission Implementing Regulation (EU) 2024/2690 laying down rules for the application of Directive (EU) 2022/2555 as regards cybersecurity risk-management measures ↗
According to the guidance, reusable evidence for third-party remote services and the yardstick for the concept of becoming aware under the reporting obligations (paras 207, 212)
European Parliament and Council of the EU · 2022
Regulation (EU) 2022/2554 on digital operational resilience for the financial sector (DORA) ↗
DORA evidence as reusable proof for third-party remote services (para. 207); possible topic of further guidance (para. 9)
European Parliament and Council of the EU · 2024
Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act) ↗
Named by the Commission as a possible topic of further guidance on the interplay with the CRA (para. 9)
European Parliament and Council of the EU · 2019
Regulation (EU) 2019/881 (Cybersecurity Act) ↗
Basis of European cybersecurity certification schemes — relevant for conformity assessment, presumption of conformity and evidence (paras 148, 153, 207)
European Commission · 2022
Commission Delegated Regulation (EU) 2022/30 supplementing the Radio Equipment Directive 2014/53/EU ↗
Existing EU type-examination certificates remain valid for CRA purposes for the cybersecurity risks covered until 11 June 2028 at the latest (paras 247–254)
European Parliament and Council of the EU · 2023
Regulation (EU) 2023/1230 on machinery (Machinery Regulation) ↗
Fully applicable from 20 January 2027; certificates remain valid for CRA purposes for the cybersecurity risks covered until 11 June 2028 at the latest (paras 255–257)
European Parliament and Council of the EU · 2019
Regulation (EU) 2019/2144 on type-approval requirements for motor vehicles as regards their general safety ↗
Vehicles and components designed and constructed exclusively for them are excluded from the CRA, generic components are not (paras 241–246)
European Parliament and Council of the EU · 2024
Directive (EU) 2024/2853 on liability for defective products (Product Liability Directive) ↗
According to the guidance, computer code itself is not a product under product liability law; liability under national tort law remains possible (footnote 9)
Planning CRA implementation on the basis of the guidance?
VamiSec supports manufacturers with scope and role clarification, cybersecurity risk assessment and threat modelling, classification by core functionality, reporting processes and readiness for the single reporting platform, as well as with product security and IoT pentests — as implementation support, not as conformity assessment or legal advice.