On August 2nd 2026, the European Union’s AI Act grace period ends, and the European Commission gains the power to supervise and fine companies that build the most powerful AI models. Prohibitions on certain AI practices took effect in February 2025, and duties for general-purpose model makers began in August 2025. The Act splits general-purpose AI into two tiers. In the lighter tier, every provider must publish a summary of training data, updated every six months. The heavier tier applies to models judged to carry “systemic risk,” defined as training using more than 10^25 floating-point operations. Providers of systemic-risk models must test, assess risks, report serious incidents, and keep systems secure. The Commission says such models are built by a handful of companies, with one industry guide estimating around a dozen global models. For non-compliance, the Commission can fine a provider up to 3% of global annual turnover or 15 million euros, whichever is higher. The AI Office will gain three significant authorities: requesting documents, running evaluations with access to source code, and demanding action up to pulling a model off the market.