The European Commission has published draft guidelines on classifying high-risk AI systems under Article 6 of the EU AI Act and opened a targeted consultation for providers, developers, businesses, public authorities, academia and citizens. Feedback closes on 23 July 2026. The guidance follows agreement on the AI Omnibus, which the European Parliament backed on 16 June and is expected to be enacted in the coming weeks.
The draft takes a broad view of the high-risk test for AI systems integrated into regulated products, or regulated products in their own right, under EU harmonisation legislation. The Commission treats enhanced regulatory scrutiny as the decisive factor, including cases where harmonised standards are required in a conformity assessment procedure, even if only internal controls are used. That approach could bring more systems in sectors such as machinery, medical devices, radio equipment, automotive and toys into scope.
For Annex III systems, Article 6(3) offers only a limited filter where AI does not materially influence decision-making and is unavailable for profiling of natural persons under data protection rules. Providers must still document a self-assessment and register qualifying systems in the EU database. If the AI Omnibus enters into force before 2 August 2026, obligations apply from 2 December 2027 for Annex III systems and 2 August 2028 for AI embedded in safety-regulated products.
