The EU AI Act, the world's first all-encompassing piece of AI legislation, was approved by the Council of Ministers this morning, which means it is set to be formally signed and then enter into force very soon, most likely in the next few weeks. The Act introduces significant regulatory requirements for AI providers, deployers and distributors, adopting a risk-based approach for AI systems and a separate regime for general purpose AI (GPAI). Key features include prohibiting certain AI uses and imposing stringent regulations on high-risk AI systems. Compliance will be driven by the introduction of official standards, with a focus on data governance, safety, transparency and systematic enforcement.
This final vote signals that the Act will shortly be formally signed and then entered into the Official Journal of the EU (OJEU). It will then come into force 20 days after its publication in the OJEU, although the majority of its provisions, including those regulating high-risk AI, will not come into force until twenty-four months after that date.
We at Bristows anticipate that the Act will influence global AI regulatory standards, though its full impact will depend on practical implementation and the development of industry and technical standards. We’re working closely with clients on how they will comply with the Act and, as importantly, develop internal governance systems and practices that will enable them to comply on an ongoing basis as they develop new AI products and services and put them out into the EU market.
This vote coincides with wider international efforts, such as the upcoming AI Safety Summit in Seoul, reflecting a growing global consensus on the need for responsible AI development and usage. It also comes as the UK government announced that it had obtained commitments from several major AI companies regarding a set of safety outcomes, so-called ‘Frontier AI Safety Commitments’, building on the Bletchley declaration from last year.