The EU's AI Act became law in August 2024 and has been switching on in stages. The part most of us will notice arrived on 2 August 2026.
What you should now be told
A chatbot must tell you it is AI at the start, unless that is obvious. AI-made images, audio and video that could pass as real must be labelled. AI-written text published to inform the public must say so, unless a person reviewed it and takes responsibility. Fines for breaking these rules reach 15 million euros or 3% of worldwide turnover. The EU has published optional icons that sites can use as the label.
What is banned
Since February 2025, AI that manipulates people in harmful ways, exploits vulnerable people, scores citizens socially, scrapes faces in bulk or reads emotions at work and school has been banned. From 2 December 2026 tools that make sexual deepfakes of people without consent, or child abuse images, are banned too.
What was delayed
An amendment called the AI Omnibus, in force since 27 July 2026, pushed the strict rules for AI used in hiring, education, credit and other high-risk areas back to 2 December 2027, and to 2 August 2028 for AI built into regulated products.
The UK and the US
The UK has no AI Act. Existing regulators such as the ICO and Ofcom apply five principles, and a parliamentary committee called for an AI bill in September 2026. The US has no federal AI law; states are acting alone. Texas's law has applied since January 2026, California now requires companion chatbots to say they are AI, and Colorado's new law gives people a right to ask for human review of important automated decisions from January 2027.
On this site
Our Ask box is labelled as AI before you use it, the pictures and clips in our hero are labelled as AI-made, and articles like this one say how they were written.