EU begins enforcing AI Act transparency rules: chatbots must say they’re AI
As of August 2, the European Commission is actively enforcing AI Act transparency requirements: interactive AI systems must disclose they aren't human, and deepfakes need machine-readable labels, with over 180 organisations already signed onto a voluntary code of practice.
The European Commission began actively enforcing a new set of AI Act transparency requirements on August 2, adding real teeth to rules that had previously existed mostly on paper. Chatbots and other interactive AI systems operating in the EU must now clearly disclose that users are talking to AI rather than a human, and any AI-generated or AI-altered content — most notably deepfakes — must carry machine-readable marks that allow it to be identified as synthetic.
The Commission’s AI Office, working alongside national authorities across member states, is responsible for enforcement, with the stated goal of reducing “deception and manipulation” and helping people make informed choices about the content and systems they interact with online. More than 180 organisations have already signed onto a voluntary transparency code of practice ahead of the harder enforcement phase, suggesting at least partial industry buy-in rather than universal resistance.
The rules represent one of the more concrete, user-facing pieces of the EU AI Act to take effect so far, in contrast to some of the Act’s higher-risk-system provisions that remain in earlier stages of phased implementation — and give a first real signal of how seriously Brussels intends to enforce the rulebook it spent years negotiating.