The European Commission published draft guidelines on 19 May 2026 on the classification of high-risk AI systems under Article 6 of Regulation (EU) 2024/1689. Publication follows an obligation under Article 96 of the AI Act, which required the Commission to issue classification guidelines no later than two years after the regulation entered into force. A targeted consultation is open until 23 June 2026; the Commission intends to incorporate feedback before adopting the final guidelines.
Article 6 establishes two separate pathways to high-risk classification. Under Article 6(1), an AI system is high-risk if it serves as a safety component of an Annex I product, or is itself such a product, and the product requires third-party conformity assessment under the relevant EU harmonisation legislation. Under Article 6(2), an AI system is high-risk if it falls within one of the use-case areas listed in Annex III, covering eight domains: biometric identification, critical infrastructure management, education, employment, access to essential private and public services, law enforcement, migration management, and administration of justice. The draft guidelines confirm that the classification is use-specific: the same AI system may be high-risk in one deployment and not in another.
Providers deploying AI systems in the Annex III domains must determine, per the draft guidelines, whether the specific use case within those domains meets the classification threshold. Providers of employment screening tools, banks using AI in credit decisions, AI deployers in healthcare, and operators of educational assessment platforms all face direct impact if a system meets the classification criteria. Those that conclude a system is not high-risk under Article 6(2) must document that determination and make it available to market surveillance authorities on request.
The draft guidelines acknowledge that an AI system performing only a narrow procedural task, or reproducing a previously human-made decision without adding a substantive new assessment, may fall outside the high-risk classification if it does not materially influence the consequential outcome. The Commission notes this determination is fact-specific and depends on the actual role the system plays in the decision chain. Application of Article 6(2) is now delayed until 2 December 2027 under the Digital Omnibus agreement of 7 May 2026, giving providers time to align their classification processes with the final guidelines before obligations apply.
Licentium assists clients in AI system classification analysis, preparation of Article 6(2) self-assessment documentation, and engagement with market surveillance authorities. Our team advises AI providers and deployers across the EU on conformity obligations. Work we undertake includes EU AI Act classification advice, provider and deployer obligation mapping, high-risk AI conformity assessment preparation, and technical documentation review.