EU AI Act

general-purpose AI obligations under the Act became enforceable on 2 August 2026, requiring providers such as OpenAI, Google DeepMind and Anthropic to meet technical documentation, capability evaluation and incident reporting duties, backed by fines up to 7% of global turnover.

Brussels' own machinery was not ready when the clock started: its enforcement-contact register still carried a September 2025 update date two days after fining power began, with Denmark, Finland and Hungary's rows blank. A month later, six of the bloc's 27 member states, Austria, Belgium, Bulgaria, Croatia, Hungary and Slovakia, still had no named regulator on the Commission's own register.

the Commission finalised its AI content-marking Code of Practice on 10 June 2026, requiring AI-generated content to be labelled from 2 August, with systems already deployed given until 2 December 2026 to comply. Zero EU broadcasters or media companies had signed by 15 July, when reporting first gave 22 July as the signatory-list cutoff.

That date proved wrong. A correction on 25 July fixed the Commission's own guidance page at 27 July, 18:00 Central European Summer Time, five days before Article 50 became legally binding. Even the law's own Deadline-tracking struggled to keep pace with the law itself.

A provisional Digital Omnibus deal reached on 7 May 2026 dropped the binding employer AI-literacy duty entirely, replacing it with a non-enforceable instruction for member states to encourage AI literacy, and pushed high-risk employment AI obligations under Annex III from August 2026 to December 2027.

The Council gave the package final approval on 29 June, confirming employers must only support staff understanding of workplace AI rather than guarantee it. The 16-month delay landed as AI-attributed job cuts kept climbing, leaving workers subject to AI-driven hiring screens without EU-level recourse for well over a year.