European Union Artificial Intelligence Act (EU AI Act)

Adopted in 2024, the Act applies in stages and reaches providers and users of AI placed on or used in the EU market, including firms based elsewhere. It sorts AI into tiers. Prohibited practices, such as certain manipulative or exploitative systems, social scoring of people and emotion inference in workplaces, are banned. High-risk systems, for example those used in hiring, credit scoring or essential services, must meet requirements on risk management, data quality, documentation, human oversight and accuracy before sale. Limited-risk systems, such as chatbots, carry transparency duties, and minimal-risk uses are left alone. General-purpose AI models face their own obligations. Penalties can reach a percentage of worldwide turnover. Legal teams use the Act's categories to decide, product by product, whether a launch needs a full conformity file or can go ahead as it is.

A software firm outside the EU sells a CV-screening tool to employers in several member states. Because recruitment is a listed high-risk use, the firm must run a risk management process, keep technical documentation, log decisions and allow human oversight. It also finds that one planned feature, inferring candidates' emotions from video, is a banned practice, and removes it before launch.