Article 11 is the central compliance artifact of the EU AI Act. It demands that providers of high-risk AI systems draw up technical documentation before placing the system on the market. Here are the 9 critical sections you must document to survive a conformity assessment.
The intended purpose, the person responsible, and the version of the AI system. Think of this as the master manifest for your model.
Detailed documentation of the logic, algorithms, and key design decisions. You must explain why you chose your architecture.
Comprehensive logs of what data was used, where it came from, how it was cleaned, and how you mitigated inherent biases (Article 10).
If you use a base model from OpenAI or Anthropic, you must document the supply chain and outline their shared responsibilities.
How can a human intervene, override, or shut down the AI system? Document the UI and UX flows for the human-in-the-loop.
The exact evaluation metrics, benchmark results, and testing frameworks used to prove the system performs safely.
Defenses against data poisoning, model evasion, and other adversarial attacks. Must map to ISO 27001 or similar standards.
Your exact plan to collect telemetry data and log incidents after the system is deployed to production.
The user manual provided to the deployer, clearly stating the system's capabilities, limitations, and known risks.
Our platform hooks into your CI/CD pipeline to generate this 40-page documentation artifact automatically on every commit.
Explore Article 11 Packages