The EU AI Act, in plain English
This EU law sets rules for AI based on how it is made and used. It can apply outside the EU when an AI tool or its results reach people in the EU.
Five risk classifications
Social scoring, manipulative practices, untargeted face scraping. Banned outright.
Hiring, credit scoring, critical infrastructure, education, law enforcement. Full conformity assessment required.
Chatbots, deepfakes, emotion recognition. Transparency obligations.
Spam filters, AI in video games. Voluntary codes of conduct.
Foundation models. Documentation, copyright policy, and (above thresholds) systemic risk controls.
Key dates
Provider vs deployer
You make an AI system or sell it under your name. You may have added safety and record duties.
You use an AI system at work. You may need to tell people, provide human review, and watch how it is used.
Check one AI tool
Answer a short set of questions and review the suggested risk level.
Start your assessment