BRUSSELS — Starting Sunday, the European Union assumes full enforcement powers under its landmark Artificial Intelligence (AI) Act, introducing the world’s most comprehensive digital rulebook for emerging technologies. Grounded in a risk-based approach, the new regulations impose strict obligations on companies depending on the potential threat AI poses to public health and fundamental rights. Under the framework, chatbots must explicitly disclose their AI identity, deepfakes and artificial content must be clearly labeled, and practices such as predictive policing, workplace emotion recognition, and behavioral manipulation are strictly banned. Existing systems have until December 2 to comply, while creative and satirical works receive exemptions.
To oversee enforcement, the EU’s AI Office—comprising legal, economic, and technical experts—gains direct authority to test advanced general-purpose models, demand market access, and restrict deployments if safety concerns are ignored. Penalties for non-compliance are severe: companies engaging in banned practices face fines of up to 7 percent of annual global turnover or €35 million, whichever is higher, with other breaches carrying fines up to 3 percent or €15 million. Additional restrictions, including bans on sexualized deepfakes, will roll out later this year, while rules for standalone high-risk systems are scheduled to take full effect by December 2027 and August 2028.
