High-risk AI systems

Classification first

Check Article 6 and the relevant Annex. Capture the system’s intended purpose and the reasoning behind the assessment. Seek qualified review for borderline applications.

Provider control areas

For applicable high-risk systems, the Act sets requirements across risk management, data and data governance, documentation, record-keeping, transparency to deployers, human oversight, accuracy, robustness and cybersecurity.

• Design and test controls against the intended use
• Maintain technical documentation and logs where required
• Plan the conformity route and post-market processes

Check the applicable date

Following the 2026 AI Omnibus amendments, Annex III high-risk rules are scheduled for 2 December 2027 and Annex I product-linked high-risk rules for 2 August 2028. Other provisions may already apply.