EU starts new rules for AI chatbots and deepfakes

New rules about artificial intelligence, or AI, started in the European Union on August 2, 2026. The European Union, or EU, is a group of countries in Europe. The rules come from a law called the AI Act. The European Commission, the office that runs the daily work of the EU, published guidelines that explain the rules.

The rules are about being honest with people. When a person talks with a chatbot, the company must tell the person that they are talking to AI, not to a human. Companies must also put a special mark on pictures, sounds, and videos that AI makes. Computers can read this mark and see that AI made the content. Companies have until December 2, 2026 to start this marking.
The rules also cover deepfakes. A deepfake is a picture, sound, or video that AI made or changed. It looks like a real person, place, thing, group, or event, and people may wrongly think it is true. When a company shows a deepfake, it must tell people that it is a deepfake.
There are more rules. Some AI systems try to read people's feelings, or they sort people by their face or body. If a company uses a system like this, it must tell the people. And when AI writes text about things that are important to the public, the text must have a label that says AI wrote it. But if a human checked and edited the text, no label is needed. Content that was made before August 2, 2026 does not need a label.
Companies that break the rules may have to pay a large fine. The fine can be up to 15 million euros, or 3 percent of the money the company made around the world in the last year. Small and medium-sized companies will be treated in a way that fits their size.
Government offices in each EU country will check that companies follow the rules. The AI Office of the EU will have only a small role. For the EU's own organizations, the European Data Protection Supervisor will check the rules.
Glossary
- 351 words
- 25 sentences
- 3 mins to read




