DevFest Nairobi 2026: Valeri Milke on securing AI agents with threat modeling and agentic AI red teaming
October 18, 2026
DevFest Nairobi 2026: our session “Securing Agentic AI” has been accepted
Session accepted: on 14 November at DevFest Nairobi 2026, Valeri Milke will show how threat modeling and agentic AI red teaming help secure AI agents. The community-led developer conference is organised by GDG Nairobi (Google Developer Groups), and its programme team has selected our talk “Securing Agentic AI: Threat Modeling & Agentic AI Red Teaming in Practice”. Valeri Milke is founder and CEO of VamiSec GmbH and a Lead Auditor for ISO/IEC 27001 and ISO/IEC 42001. The event takes place on Saturday, 14 November 2026, at the Sarit Expo Centre on Karuna Road in Nairobi, Kenya.
“Build. Secure. Scale.” — DevFest Nairobi 2026, hosted by GDG Nairobi
DevFests are community-led developer conferences held around the world, each organised independently by a local Google Developer Group. In Kenya's capital, GDG Nairobi is hosting DevFest 2026. The motto “Build. Secure. Scale.” sets the direction: topics range from AI and agentic applications through cloud, web and mobile to security. The event is aimed at developers, founders, builders, tech leaders and communities. According to GDG Nairobi, attendance is free but places are limited; all registration details are on the official event page at https://gdg.community.dev/events/details/google-gdg-nairobi-presents-devfest-nairobi-2026/.
Build, Secure, Scale: three tracks with a local focus
The call for papers grouped submissions into three tracks: Build stands for the craft, Secure for trust and security, and Scale for the ambition to turn a prototype built in Nairobi into something that serves an entire region. Submissions for Secure and Scale were especially welcome. For Secure, the call lists topics such as securing autonomous agents in production, prompt injection and how to defend against it, as well as guardrails, human-in-the-loop and safe failure modes. There is also a clear local focus: talks should take into account the realities of African networks, budgets, devices and markets. And DevFest Nairobi sees itself as a demo-led conference — no slides without proof.
Agentic AI security: what changes when AI agents use tools
Agentic AI security becomes very concrete as soon as a model can call tools, access data and trigger real actions. Weaknesses in how external content, permissions or trust boundaries are handled then reach all the way into the connected systems. That is why the security question is not only what a model says, but what an agent is allowed to do — and on whose behalf. A typical pattern: an indirect prompt injection, hidden in seemingly harmless content, is delegated as a task to another agent and ends in the misuse of a tool. Effective safeguards are above all architectural: least privilege, authorisation enforced in the backend and approvals bound to the specific tool and its parameters.
The session: threat modeling for AI agents meets agentic AI red teaming
The session centres on three steps that development teams can put into practice straight away: first, understand the system — its assets, permissions and trust boundaries. Second, identify realistic attack paths across agents, tools and connected services instead of listing risks in the abstract. Third, use red teaming in a targeted way to challenge security assumptions and evaluate the intended controls. The aim is to turn security questions into concrete engineering decisions and to build reliable evidence that the controls work. In line with the DevFest motto, the focus is on AI systems with security considered from the start — not left until shortly before go-live.
The Agentic Era: the overarching theme of the 2026 DevFest season
The Nairobi motto does not stand alone: worldwide, the 2026 DevFest season runs under the overarching theme “Build, Secure, Scale: Developers and Builders in the Agentic Era”, announced by Google in September, although each GDG sets its own agenda. The topics listed on GDG Nairobi's event page include AI with Gemini and Gemma, Google Cloud, Firebase, Android and Flutter, as well as web, UX and hands-on workshops. Anyone using these platforms to build agents that call tools and process data faces exactly the questions our session addresses: which permissions does an agent really need — and how can you prove that its boundaries hold?
The practice behind it: AI threat modeling with MAESTRO, agentic AI pentesting and red teaming
What we will present in Nairobi is everyday project work for us: we model agentic systems using MAESTRO, combine STRIDE with the OWASP catalogues and prioritise risks based on the real architecture; through agentic AI pentesting and red teaming, we then check which of the risks in the threat model can actually be exploited. MAESTRO, introduced by the Cloud Security Alliance (CSA) in February 2025, extends established methods such as STRIDE to cover the architectural layers of agentic systems. We also draw on the OWASP Top 10 for Agentic Applications (2026 edition), the OWASP Top 10 for LLM Applications 2026 and the OWASP AI Testing Guide, among others. Find out more at https://vamisec.com/en/ai-threat-modeling-maestro and https://vamisec.com/en/agentic-ai-pentesting. VamiSec is itself certified to ISO/IEC 27001.
Meeting the developer community in Nairobi — and beyond
We look forward to exchanging ideas with the developer community on site. If you would like to talk about agentic AI security, threat modeling or AI red teaming — in Nairobi or online — you can reach us at contact@vamisec.com. We will add the exact time slot of our session here as soon as it is published. Six days after DevFest, on 20 November 2026, Valeri Milke will also speak at the Agentic AI Foundation's MCP Dev Summit Nairobi on “Threat Modeling Agentic AI: MCP + MAESTRO + STRIDE”. Can't make it to Nairobi? Our knowledge pages cover the fundamentals of AI red teaming at https://vamisec.com/en/wissen/agentic-ai-security/ai-red-teaming and of the OWASP Top 10 for Agentic Applications at https://vamisec.com/en/wissen/agentic-ai-security/owasp-agentic-top-10.
All links for the event
Official event page & registration (free, limited places): https://gdg.community.dev/events/details/google-gdg-nairobi-presents-devfest-nairobi-2026/ · GDG Nairobi: https://gdg.community.dev/gdg-nairobi/ · DevFest worldwide: https://developers.google.com/community/devfest · MCP Dev Summit Nairobi schedule: https://events.linuxfoundation.org/mcp-dev-summit-nairobi/program/schedule/ · AI threat modeling with MAESTRO at VamiSec: https://vamisec.com/en/ai-threat-modeling-maestro · Agentic AI pentesting at VamiSec: https://vamisec.com/en/agentic-ai-pentesting · Threat-Model-Guided Security Testing (ICSPIS 2026): https://vamisec.com/en/wissen/agentic-ai-security/threat-model-guided-security-testing · Agentic AI security at VamiSec: https://vamisec.com/en/agentic-ai-security
Do you have questions about your organization's IT security?
Free Initial Consultation →