The European Union has adopted what it characterizes as the world’s first comprehensive law on artificial intelligence, which bans some uses of the technology while regulating others. The EU AI Act takes effect Aug. 1, although the beginning of enforcement for different parts of the law will be spread across the next six to 36 months.
The EU AI Act divides AI technology by risk level, according to a summary of the law by the European Parliament. “Unacceptable risk” AI technologies are systems deemed to pose a threat to people, and will be banned. Examples include systems used for cognitive behavioral manipulation of people or specific vulnerable groups; classifying people based on behavior, socio-economic status or personal characteristics; and using biometric identification. The ban on “unacceptable risk” AI begins six months after the law takes effect.
“High risk” AI technologies “that negatively affect safety or fundamental rights” will be placed into two categories. The first category is AI used in products falling under the EU’s product safety legislation, such as toys and cars. The second category is AI falling into specific areas that must be registered in an EU database, such as systems for managing and operating critical infrastructure. High-risk systems will be given 36 months to comply.
Generative AI technologies, such as ChatGPT, are not considered high risk but still must comply with transparency requirements and EU copyright law, although more advanced AI models that may pose systemic risk will be subject to evaluations and must report “serious incidents,” according to the European Parliament.
