Deepfake videos of public figures, AI chatbots posing as humans, machine-generated news text with no editorial oversight — as of today, all of that requires a disclosure label in the EU. The European Commission has announced that the AI Act’s transparency obligations took effect on August 2, 2026, marking the most consequential enforcement milestone the regulation has hit so far.
The rules apply to two broad areas. First, AI-generated or manipulated content — including deepfakes of real people, emotion recognition outputs, and AI-written public interest text with no human review — must carry clear, visible labels plus machine-readable marks. The EU has published a set of official icons for this. Second, anyone deploying a chatbot, AI agent, or avatar must tell users upfront they are not talking to a human. No burying it in terms of service.
Enforcement sits with national market surveillance authorities, the European AI Office, and the European Data Protection Supervisor. Penalties are real: up to €15 million or 3% of global annual turnover for companies, and up to €750,000 for EU institutions. SMEs get some proportionality consideration, but the framework is not soft.
For developers and founders building in Europe, or shipping products to European users, this changes the compliance checklist immediately. Labeling pipelines, disclosure flows, and user interface copy all need review. The Commission has also published guidelines explaining how to demonstrate compliance, including through a code of practice. That’s worth reading before your legal team does.
This is the third phase of the AI Act rolling out since it entered force in August 2024. The high-risk system requirements still come later. But the transparency layer is here now, and regulators are watching.
