EU AI Act
Artificial Intelligence Act
An EU law that sets rules for developing and using AI based on how risky it is.
In Simple Terms
The EU AI Act is a set of European rules that decide how AI can be used and what standards it has to meet, depending on how risky it is, so that AI doesn't threaten people's rights or safety. The most dangerous uses are banned outright. The next riskiest AI has to meet duties like safety checks and record-keeping, and chatbots and generative AI have to make it clear that they're AI. Companies that don't follow the rules face heavy fines.
Behind the Name
The name is a plain description of what it is: a law (an "Act") about AI (artificial intelligence) made by the EU (the European Union, a group of European countries working together). The official name in English is the Artificial Intelligence Act, and it's sometimes shortened to "AI Act". Putting "EU" in front just tells you whose law it is.
Take a Closer Look!
The EU AI Act is a law that sets standards for AI developed or used in the EU, with the goal of protecting people's safety and human rights.
Its key feature is that it sorts AI into 4 risk levels based on how much harm it could cause, with different rules for each level.
At the top of the scale, AI that's too dangerous, such as systems that manipulate people's behavior, is banned entirely. High-risk AI, such as AI used in medical devices or hiring decisions, comes with strict duties like safety checks during development and keeping records.
AI that generates text or images, meanwhile, has to be transparent: its output must be machine-detectable as AI-made, and users must be told that it's AI in certain situations.
The law doesn't stop at Europe's borders. It also applies to companies around the world that offer AI services to the European market. So any company providing AI in the EU needs to meet the standards that apply to its kind of AI.
Serious violations can bring hefty fines.