The European Commission finalised its Code of Practice on marking and labelling of AI-generated content on Wednesday 10 June, after the closing plenary1. The Code is a voluntary framework signatories can use to show compliance with the EU AI Act, the bloc's law governing artificial intelligence. It now goes through an adequacy assessment by the Commission and the AI Board, the EU body overseeing the Act, and applies from 2 August 2026, 46 days away for new systems. Deployers must label AI-generated text on matters of public interest, unless that text sits under human editorial responsibility2.
The finalisation completes a marking framework that has been assembling for months. The EU AI Omnibus gave systems already on the market until 2 December to carry machine-readable marks, and Spotify adopted the DDEX metadata standard, an industry format for tagging tracks, to flag AI-generated music ahead of the rules. Those earlier moves were anticipating exactly this rulebook, and both bought the affected companies a softer landing than the new-system deadline allows.
The Code exempts text produced under human editorial sign-off from the labelling duty, the clause that brings the framework into a working newsroom. A broadcaster's compliance posture turns on where it draws the line between machine output and human responsibility. Sign-up is open and no media company has yet signed, which is itself a leading indicator: in this beat, Code signatories versus non-signatories separate the companies that have built synthetic-content governance from those still deciding. France Televisions naming its vendors is one reading of how an early mover prepares; an unsigned register five weeks before the deadline is the other.
