On August 2, 2025, the obligations for providers of general-purpose AI (GPAI) models under the European Union’s AI Act began to apply. The European Commission’s own page on the regulatory framework states that “the AI Act rules on GPAI became effective in August 2025.” That start date still stands, but the originally planned full-application date of 2 August 2026 no longer holds: under the “Digital Omnibus” amendments (Commission proposal adopted 19 November 2025, political agreement reached 7 May 2026), the EU deferred the bulk of the high-risk obligations, with rules for high-risk systems in certain areas now applying from 2 December 2027 and product-embedded high-risk systems from 2 August 2028. The GPAI rules and the prohibitions, which took effect earlier, are not pushed back by the deferral. See 2026-eu-ai-act-omnibus-deferral.
GPAI providers face obligations such as maintaining technical documentation, putting in place a policy to comply with EU copyright law, and publishing a detailed summary of the content used to train their models. To support compliance, the Commission released guidelines and a voluntary code of practice in July 2025, describing the aim as “a clear and actionable framework for providers of GPAI models to comply with the AI Act, reducing administrative burden, and fostering innovation while safeguarding fundamental rights and public trust.” The Commission’s direct enforcement powers over GPAI providers were set to apply from August 2, 2026.
This was a milestone because it marked the first time a major jurisdiction placed binding, foundation-model-specific obligations on the companies that build frontier AI, rather than only on those who deploy it. It set an early template for AI governance that providers worldwide had to account for in how they document, train, and release models.