EU turns AI law into rules

The European Commission has started writing the rulebook for enforcing the European Union’s Artificial Intelligence Act, opening consultations on how investigations, emergency orders and fines should work. (techpolicy.press) The consultations cover procedural rules for the Act, including how authorities could impose interim measures before a case is finished and how long they would have to pursue penalties. One proposal sets a five-year limitation period for fines. (techpolicy.press) The Artificial Intelligence Act itself entered into force on August 1, 2024, after the European Union adopted it as the bloc’s first broad law for artificial intelligence systems. Its obligations phase in over time rather than all at once. (eur-lex.europa.eu) That phased rollout already started. Rules on prohibited artificial intelligence practices began applying on February 2, 2025, while obligations for general-purpose artificial intelligence models and later deadlines for high-risk systems follow on later dates set in the law. (eur-lex.europa.eu) The new consultations move the law from broad categories into case-handling mechanics. Companies can read the headline ban on a system, but enforcement turns on details such as deadlines, evidence rules and when regulators can act before a final decision. (techpolicy.press) That matters for providers of foundation models and other artificial intelligence systems selling into Europe, because the Act applies based on where systems are placed on the market or used, not just where a company is based. The law also allows fines that can reach fixed euro amounts or percentages of global annual turnover, depending on the violation. (eur-lex.europa.eu) European Union officials have been building the rest of the enforcement structure in parallel. The European Commission set up an Artificial Intelligence Office in 2024 to help supervise general-purpose artificial intelligence and coordinate implementation across the bloc. (commission.europa.eu) Industry groups have pushed for clearer guidance before penalties bite, especially for general-purpose artificial intelligence providers facing new transparency, safety and documentation duties. Civil society groups, meanwhile, have argued that weak enforcement procedures could blunt a law sold as a check on harmful uses of artificial intelligence. (techpolicy.press) The Commission’s consultations are the part where Europe decides how fast, how hard and how consistently the Artificial Intelligence Act will be enforced. After the headline law, this is the stage that tells companies what regulators can actually do. (techpolicy.press)