EU Trilogue Reaches Partial Agreement on AI Act High-Risk Rules
European negotiators agree on classification rules for high-risk AI systems under the draft AI Act, but several contentious issues remain unresolved. The partial deal includes exemptions for systems performing purely accessory tasks.
European negotiators reach a partial agreement on Article 6 of the draft EU AI Act, establishing classification rules for high-risk artificial intelligence systems. The trilogue session involving the European Parliament, the European Council, and the European Commission concludes with a consensus on a certification regime for AI systems deployed in high-risk scenarios.
The agreed text introduces an exemption that prevents certain AI systems from being classified as high risk if they perform purely accessory tasks. To qualify for this exemption, an AI system must execute a narrow procedural task, detect deviations in decision-making patterns, avoid influencing major decisions like loan approvals or job offers, or simply improve work quality like a smart grammar checker.
Despite this progress, consumer and privacy activists express concern about allowing companies to self-determine whether their AI systems fall under the high-risk category. Significant questions remain unresolved regarding how organizations prove their systems meet the exemption criteria and what specific burden of proof is required to avoid strict regulatory oversight.