From the journal

European Commission Publishes Draft Guidelines on High-Risk AI System Classification, June 2026

The European Commission published draft guidelines to help providers and deployers of artificial intelligence systems determine whether a given system qualifies as high-risk under Annex III of Regulation (EU) 2024/1689, the EU AI Act. A targeted industry consultation was open until 23 June 2026, with the deadline extended to 23 July 2026. The Commission plans to adopt final guidelines by the end of 2026.

3 min read

The European Commission published draft guidelines clarifying how providers and deployers of artificial intelligence systems should determine whether a given system qualifies as high-risk under Regulation (EU) 2024/1689, the EU AI Act. The draft was open for targeted industry consultation, with the original deadline of 23 June 2026 extended to 23 July 2026. Final guidelines are expected by the end of 2026.

Under Article 6 and Annex III of the EU AI Act, an AI system qualifies as high-risk if it satisfies one of two criteria. First, it functions as a safety component of a product subject to EU harmonisation legislation that requires third-party conformity assessment. Second, it falls within one of the use cases listed in Annex III, covering eight sectors: biometric identification and categorisation of natural persons; management of critical infrastructure; education and vocational training; employment and workers management; access to essential private and public services and benefits; law enforcement; migration and border control management; and administration of justice and democratic processes. The draft guidelines supply practical examples for each Annex III category to assist providers in making the classification assessment.

Providers that classify an AI system as high-risk must complete a conformity assessment before placing the system on the EU market, register the system in the EU database under Article 71, and meet the requirements of Articles 9 to 15, covering data governance, technical documentation, record-keeping, transparency toward deployers, human oversight, accuracy, robustness, and cybersecurity. Deployers using high-risk AI systems in employment decisions, educational assessments, or for access to essential services must conduct fundamental rights impact assessments under Article 27. Providers building AI products for any Annex III sector should treat the draft guidelines as the operative interpretation of Article 6 for current compliance planning.

The draft guidelines cover classification only, not conformity assessment procedures or enforcement. The Commission is separately preparing guidelines on Article 50 transparency obligations, on serious incident reporting, and on the full set of high-risk requirements for providers and deployers. Under Article 7, the Commission can expand Annex III by delegated act to add further high-risk use cases, so the final scope of the high-risk category may differ from the current Annex III text.

Licentium advises AI developers, technology companies, and deployers on EU AI Act compliance, including high-risk classification analyses, conformity assessment preparation, and fundamental rights impact assessments. We help clients determine whether their systems fall within Annex III scope and prepare for the applicable obligations. Contact us to discuss your EU AI Act compliance position. Work we undertake includes AI system classification assessment, conformity assessment support, AI regulatory strategy, EU AI Act implementation planning, and general-purpose AI model regulatory analysis.

Source: European Commission, Draft Guidelines on the Classification of High-Risk Artificial Intelligence Systems (EU AI Act, Annex III), consultation extended to 23 July 2026

AI Regulatory

More from the journal

See all

Google Engineer Charged with Commodities Fraud on Polymarket Using Confidential Data, May 2026

On 27 May 2026, the U.S. Attorney for the Southern District of New York unsealed a criminal complaint charging Michele Spagnuolo, a Google staff software engineer, with commodities fraud, wire fraud, and money laundering. Spagnuolo allegedly used confidential internal Google Search data to place approximately $2.75 million in bets on Polymarket event contracts tied to Google's Year in Search report between October and December 2025, netting roughly $1.2 million in profit. The CFTC filed a parallel civil action seeking penalties and trading bans.

Georgia Enacts Payment Stablecoin Act, Establishing Issuer Licensing Regime, May 2026

On 11 May 2026, Georgia signed HB 1272 (Act 452), the Georgia Payment Stablecoin Act, into law. The statute directs the Georgia Department of Banking and Finance to license stablecoin issuers incorporated under Georgia or foreign law. Licensed issuers must maintain one-to-one reserves of eligible assets and may only engage in stablecoin issuance, redemption, reserve management, and related custodial activities. The law takes effect on the earlier of 18 January 2027 or 120 days after federal GENIUS Act implementing regulations are finalized.

European Commission Opens Targeted Consultation on MiCA Review, May 2026

On 20 May 2026, the European Commission launched a targeted consultation on Regulation (EU) 2023/1114, the Markets in Crypto-Assets Regulation (MiCA), running alongside a parallel public consultation. The targeted consultation spans 86 questions across four thematic blocks and invites responses from industry representatives and public authorities. Submissions close 31 August 2026, with results feeding into the Commission's review reports under Articles 140 and 142 of MiCA.

Ready to launch without the regulatory guesswork?

Book a 30-minute consultation. We'll map your AI or licensing path and tell you exactly what's required, in plain language.