The European Commission released its final Article 50 transparency guidelines on July 20, 2026, giving providers of chatbots, deepfake tools, and generative content systems 13 days to comply before the AI Act’s August 2 deadline. The 51-page instrument lands so close to the trigger date that most legal teams will read it and the regulation into force in the same working fortnight.

Fines run to €15 million or 3% of worldwide annual turnover, whichever is higher, and enforcement flows through national market surveillance authorities across all 27 member states. Executive Vice-President Henna Virkkunen framed the guidelines as intended to “support the smooth and effective application of the AI Act.” That’s the official register. The operational reality is a watermarking mandate the underlying technology can’t yet reliably meet.

The document is non-binding. Only the CJEU can give authoritative interpretation, as Bird & Bird notes, but the AI Office and national regulators will treat it as the primary Article 50 reference, which is the same thing with extra steps.

Several late changes matter. Deepfakes generated before August 2 don’t need retroactive labeling; the text-content trigger keys off publication date, not generation date; and “persons” now expressly covers digital replicas, realistic avatars, and voice or performance likenesses. Pre-existing generative systems get until December 2, 2026 to bring machine-readable marking into conformity, under the Digital Omnibus provisional agreement of May 7. The parallel Code of Practice on Transparency of AI-Generated Content, which the Commission has assessed as adequate, closed for signatures at 6 p.m. CEST on July 22. Non-signatories must now demonstrate compliance on their own.

Brussels has done this before. GDPR shipped with a two-year runway and still caught half the market flat-footed. This time the runway is 13 days.

Sources

Sources