What is the EU AI Act, and how do other places regulate AI?
A hiring tool and a spam filter are both AI, but under Europe's AI Act one faces strict rules and the other almost none.
Guess first
?
Which one is it?
The EU AI Act is a European Union that sets common rules for artificial intelligence across the EU, with stricter rules for AI that could cause more harm. It entered into force on 1 August 2024, and its provisions come into operation gradually over the following 6 to 36 months. It places duties on AI and on organisations that use AI in a professional context, rather than creating individual rights.
The Act sorts AI by risk. Unacceptable-risk uses, such as or manipulating human behaviour, are banned unless an exemption applies. High-risk uses, such as AI in hiring, education, health or law enforcement, face obligations on quality, transparency, human oversight and safety. Limited-risk systems, such as tools that generate or manipulate images, sound or video (like deepfakes), must tell people they are dealing with AI. Minimal-risk systems, such as spam filters or video games, are not regulated.
Step 1: Unacceptable risk
Banned, unless an exemption applies.
Step 2: High risk
Quality, transparency, human oversight, safety duties.
Step 3: Limited risk
Tell people they are dealing with AI.
Step 4: Minimal risk
Not regulated.
models face their own list: must publish a summary of their training data, adopt a policy to comply with copyright law and provide technical documentation. Breaking the Act's prohibitions can bring fines of up to EUR 35 million or 7% of worldwide , whichever is higher.
Elsewhere the approach is different. China's rules for generative AI include requirements to watermark generated images or videos. In July 2023, a group of companies including OpenAI, Alphabet and Meta signed a voluntary agreement with the US administration to watermark AI-generated content. Reactions to the EU Act are mixed: some startups welcomed the clarity, others fear lost competitiveness, and Amnesty International says it offers limited human rights protections.
Quiz me
0/3
Recap
Riskier AI, stricter rules: ban, regulate, label, or leave alone.
💡 A trick to remember it · Four rungs on the risk ladder: ban, check, label, leave.
Surprising fact · Its top fine is EUR 35 million or 7% of worldwide annual turnover, and it can reach companies outside the EU.
Connects to
Sources (3)
No source, no claim. Every fact in this lesson (17 claims) cites at least one of these.