Agentic AI Security: securing & governing AI systems

LLM and agentic AI pentesting, Prompt Injection, OWASP LLM Top 10, MAESTRO Threat Modeling and AI governance: how to test AI systems, make them transparent and operate them in full regulatory compliance.

View our services All knowledge categories

Agentic AI systems act instead of merely answering — with tool access, memory and their own decisions. This creates attack surfaces that classic security testing does not cover: Prompt Injection, Memory Poisoning, tool misuse and compromised Agent Skills. These knowledge pages combine offensive testing (OWASP LLM Top 10, agentic threats), engineering transparency (AI-SBOM, automated Threat Modeling) and governance (EU AI Act, ISO/IEC 42001) into one end-to-end approach. Runtime control based on the OWASP Agent Control Standard and evidence-based security testing from our ICSPIS 2026 paper make that protection verifiable.

Topics & knowledge pages

Five topic areas at a glance — each with a short overview and the matching deep dives.

Runtime controlNew

Agent Defence Strategies & Tactics

Layered defence for AI agents — from runtime control with OWASP ACS to incident response.

Protocol security

MCP Security

Securing Model Context Protocol deployments — from authorization to pentesting.

Offensive securityNew

Testing & Red Teaming

Attack your AI systems before someone else does — threat-model-guided and backed by evidence.

Threat modeling

Threat Landscape & Secure Design

Understand the AI threat landscape, model threats systematically and build AI systems secure by design.

Regulation & risk

AI Governance

EU AI Act, ISO/IEC 42001 and AI frameworks — know your obligations and steer AI risks.

Deep Dives

In-depth knowledge pages on the topics that currently raise the most questions — with verified sources and concrete measures.

ISO/IEC 27090

ISO/IEC 27090:2026 is the guidance standard from ISO/IEC JTC 1/SC 27 on security threats and compromises to AI systems, published on 14/10/2026 as a first edition of 56 pages. The standard describes nine AI-specific attack types with their consequences, detection and mitigation across the entire AI life cycle and is not certifiable. This page puts the standard in context, shows threats by life-cycle phase, maps them to ISO/IEC 27001, 42001, the AI Act, NIST and OWASP, offers a Gap Check for self-assessment and the whitepaper "ISO/IEC 27090 for CISOs".

View knowledge page

Agent Control Standard (ACS)

The OWASP Agent Control Standard (ACS) is an open specification under which a Guardian Agent checks an AI agent’s actions at defined hooks and, based on policy, allows, blocks, modifies, submits them for approval or defers them. This deep dive shows what version 0.1.0 delivers today — and where its limits lie.

View knowledge page

Threat-Model-Guided Security Testing

A shared trace only shows that events belong together. It does not show whether an AI agent respected approval, target, authorization and data provenance along the way. Our paper, accepted at ICSPIS 2026 (IEEE), shows which telemetry that requires. This page maps the results to the OWASP Agent Control Standard (ACS).

View knowledge page

OWASP APTS

AI-driven pentest platforms increasingly execute attack chains on their own — the OWASP Autonomous Penetration Testing Standard (APTS) is the first governance framework built for exactly that: 173 requirements, 8 domains, 3 compliance tiers, 4 autonomy levels.

View knowledge page

MCP Security

The Model Context Protocol connects AI applications with tools, data, and systems – creating a new attack surface in the process. We put the current guidance from the NSA, OWASP, and Wiz into context and show how you can secure MCP servers and clients.

View knowledge page

OWASP Agentic Top 10

The ten key security risks of autonomous AI agents – from Agent Goal Hijack to Rogue Agents – based on the official OWASP publication (version 2026, December 2025).