Brussels has crossed a regulatory threshold. Starting Sunday, the European Union can actively police its landmark artificial intelligence rules, backed by fines that could reach €35 million or seven percent of a company's global turnover. The enforcement marks the transition of the AI Act from a legislative blueprint into a working legal framework with real consequences. Main Developments The AI Act's risk-based approach now applies across the bloc. Systems posing higher risks to citizens' rights or health face stricter obligations, while banned uses include predictive policing, emotion recognition in workplaces or schools, and AI designed to manipulate human behavior. Chatbots must now disclose their AI identity to users, and AI-generated content such as deepfakes requires clear labelling. Existing systems have until December 2 to comply, though exemptions exist for artistic, creative, satirical, and fictional works. Read also: Why US imports surged 39% in Pakistan's FY26 trade shift The AI Office, a regulatory body attached to the EU executive, gains enforcement authority over the most advanced general-purpose models. Staffed by dozens of tech experts, lawyers, and economists, it can test chatbots and other advanced systems for compliance, with companies required to grant access to their models. In some cases, the EU can demand access before a model reaches the market. The bloc can also restrict deployment within its borders if concerns about a provider remain unaddressed. National authorities will handle enforcement for smaller AI systems. Background The AI Act was adopted in 2024 as part of the EU's comprehensive digital rulebook, which already covers social media, online retailers, and search engines. Its rules have been introduced in stages rather than all at once. This year's enforcement gap became visible when the EU struggled to gain access to American AI developer Anthropic's Mythos model. The new powers are designed to prevent such access disputes from recurring. Why It Matters For businesses, the stakes are substantial. The stiffest penalties target companies allowing banned practices, with fines up to seven percent of annual worldwide turnover or €35 million, whichever is higher. Other violations carry fines up to three percent of global revenue or €15 million, and firms can also be penalized for hindering or failing to cooperate with EU probes. The EU's approach positions it as the global reference point for AI regulation. Companies operating across borders must now weigh compliance costs against market access to one of the world's largest consumer bases. What's Next Additional AI Act provisions take effect later this year. December brings a ban on AI systems generating sexualised deepfakes, introduced after global outrage over non-consensual nudes produced by Elon Musk's chatbot Grok. Implementation of some high-risk AI rules is being delayed to give companies more preparation time. Stand-alone high-risk AI systems face rules from December 2027, while AI tools embedded in other products must comply by August 2028.