EU keeps AI Act central
Europe is keeping the AI Act as its main legal framework even as political debate grows about loosening parts of the rulebook. Analysts note the Act’s strength comes from a harmonised, single-instrument approach across all 27 member states rather than a patchwork of local laws, while Brussels has pulled back on a separate AI Liability Directive that would have created a parallel civil-liability regime. Practical guidance for implementers is converging on risk-based classification and modular compliance workstreams. (mondaq.com, mondaq.com, braintrust.dev, thenextweb.com)
Europe is still building its artificial intelligence rulebook around one law: the European Union Artificial Intelligence Act. The regulation entered into force on August 1, 2024, and Brussels has kept it as the core framework across the bloc’s 27 member states. (digital-strategy.ec.europa.eu) That matters because the Act is a regulation, not a directive. In European Union law, that means one directly applicable rulebook for the single market, rather than 27 separate national transpositions. (eur-lex.europa.eu) The rollout is staggered. Bans on prohibited uses and artificial intelligence literacy duties started applying on February 2, 2025, general-purpose artificial intelligence model obligations started on August 2, 2025, and the broad regime is scheduled to be fully applicable from August 2, 2026, with some provisions running through August 2, 2027. (digital-strategy.ec.europa.eu, ai-act-service-desk.ec.europa.eu) Brussels has also stepped back from creating a second artificial intelligence-specific liability track. The European Commission withdrew the proposed Artificial Intelligence Liability Directive in 2025 after political support fell away, leaving the Artificial Intelligence Act as the main dedicated legal instrument. (europarl.europa.eu, mondaq.com) The current debate in Brussels is not about replacing the Act with a national patchwork. It is about whether to simplify parts of implementation while preserving a central European Union framework for market access, enforcement, and cross-border trade. (commission.europa.eu, thenextweb.com) The Commission’s own simplification agenda points in that direction. A digital omnibus package presented on November 19, 2025 proposed targeted changes to artificial intelligence rules while keeping centralised oversight and delaying some high-risk obligations until support tools and standards are in place. (commission.europa.eu, commission.europa.eu) For companies, the practical work is becoming more concrete. The Act sorts systems by risk, and the first compliance question is usually which bucket a tool falls into: prohibited, high-risk, transparency-triggering, or lower-risk. (digital-strategy.ec.europa.eu, braintrust.dev) That is why legal and technical teams are increasingly organizing compliance in modules instead of one giant rewrite. Common workstreams include inventorying models, mapping use cases, checking whether a system is high-risk, documenting data and testing, and assigning owners for procurement, security, and human oversight. (braintrust.dev) Brussels has also been filling in the operating manual around the law. The General-Purpose Artificial Intelligence Code of Practice was published on July 10, 2025 as a voluntary tool to help providers meet the Act’s model-level duties, and the Commission said signatories could get streamlined compliance. (digital-strategy.ec.europa.eu, digital-strategy.ec.europa.eu) Critics of the European approach say the compliance load could slow smaller firms and weaken the bloc against United States rivals. Supporters argue that one harmonised rulebook is exactly what keeps Europe from fragmenting into conflicting national standards. (thenextweb.com, mondaq.com) So the center of gravity has held. Europe may trim deadlines, guidance, and paperwork, but its artificial intelligence policy is still being built around a single law that applies across the whole union. (digital-strategy.ec.europa.eu, commission.europa.eu)
