Current law still points to 2 August 2026 for most obligations. The 7 May political agreement is not final law yet.

Update

Draft high-risk classification guidelines are open for feedback

The Commission issued draft guidelines on 19 May 2026 to help providers and deployers assess whether an AI system is high-risk. Feedback is open until 23 June 2026.

Last reviewed June 5, 2026
Current law firstPractical, evidence-led guidanceClear next steps

What is new

The draft guidelines interpret Article 6 and include practical examples for both high-risk routes: AI used as a safety component or product under Annex I legislation, and AI systems falling into Annex III use cases such as biometrics, critical infrastructure, education, employment, migration, asylum, border control, and related sensitive areas.

Why it matters now

Even with the May 2026 Digital Omnibus political agreement pointing to later high-risk dates, classification work is still useful. Teams need to know whether a use case is likely high-risk before deciding how much evidence, vendor diligence, FRIA planning, or technical documentation work is needed.

  • Review your HR, education, credit, insurance, and biometric use cases against the draft examples.
  • Document uncertainty explicitly instead of forcing a yes/no classification too early.
  • Use the consultation window if your product category is unclear or poorly represented.

FAQ

Are the draft guidelines binding?

No. They are draft Commission guidance, but they reflect the Commission’s interpretation and will inform final guidance and enforcement practice.

Does the Digital Omnibus make classification irrelevant?

No. The timing may move, but classification still determines which obligations and evidence lanes matter.

Next step

Turn this reading into an actionable report

Use the free scanner to map your likely role, detect likely obligations, and see which evidence is missing.