The European Commission published draft guidelines on 19 May 2026 setting out its interpretation of the classification criteria for high-risk AI systems under Article 6 of the EU AI Act (Regulation (EU) 2024/1689). The guidelines are at the consultation stage; interested parties may submit input until 23 June 2026 at 22:00 CET, after which the Commission will finalise the text.
Article 6 of the EU AI Act establishes a two-tier classification mechanism. Under Article 6(2), an AI system listed in Annex III is presumed high-risk unless the provider can demonstrate it poses no significant risk to health, safety, or fundamental rights. Article 6(5) expressly mandates that the Commission issue guidelines containing practical examples. The draft published on 19 May operationalises that mandate, providing both a general interpretive analysis and a non-exhaustive catalogue of examples covering all Annex III domains.
Providers and deployers subject to the EU AI Act must assess whether their systems fall within the high-risk category before placing them on the market or putting them into service. Systems classified as high-risk trigger obligations including conformity assessments, technical documentation, registration in the EU database, and ongoing post-market monitoring. Deployers in regulated sectors - financial institutions, insurers, healthcare providers, and public authorities - face the greatest immediate exposure, as Annex III covers AI used in credit scoring, employment decisions, access to public benefits, and law enforcement.
The draft guidelines are not yet legally binding. The consultation period runs until 23 June 2026 and the Commission may revise the examples before finalisation. Providers who believe their system falls outside the high-risk categories under Article 6(3) must document that assessment in technical files. Open questions remain around AI systems that straddle multiple Annex III domains and AI components embedded in larger products already subject to sector-specific EU law.
Licentium advises clients on EU AI Act compliance obligations, including high-risk system classification, conformity assessment procedures, and technical documentation requirements. Work we undertake includes AI Act readiness assessments, high-risk classification analysis, provider and deployer compliance programme design, and engagement with national market surveillance authorities.