EU AI labelling deadline for existing systems falls on 2 December
Reviewed by Helen Jones
The EU AI Act started applying on 2 August 2026. Transparency duties under Article 50, including telling people they are interacting with AI and labelling synthetic content, were not postponed by the Digital Omnibus. A later amendment, Regulation (EU) 2026/1744, in force from 27 July 2026, gives systems already on the market extra time to add content marking. That grace period runs to 2 December 2026. High-risk rules for Annex III uses, including many employment tools, still start on 2 December 2027. Embedded Annex I systems follow on 2 August 2028. Rules on general-purpose AI and banned practices already apply.
Until this summer, many teams treated the Act as a 2027 problem for recruitment software and other high-risk workplace tools. Blank chats, custom GPTs, and Claude or ChatGPT Projects could keep producing drafts with no shared rule on disclosure, watermarks, or who checks the output. That split is now the point. Disclosure of AI interaction is already live. The next dated operator task for many existing generative systems is content marking by 2 December 2026, not the later employment rules. Classification still follows the use case, not the brand of the chatbot.
Analysis
This is a change to act on for tools already in use, not a reason to freeze hiring systems that sit on the 2027 date. Before 2 December, list every generative workflow your team already runs, then add one review step that records whether the output is labelled as AI-made and whether the user was told they were talking to a machine.
Source note
Pulse published by Collab365 Spaces, reviewed by Helen Jones on . Cite as "EU AI labelling deadline for existing systems falls on 2 December", Collab365 Spaces. 3 sources referenced.