AI enforcement advances as US framework stays opaque

AI regulation is moving into enforcement in several jurisdictions, with the EU and China taking the most concrete steps. The European Commission’s AI Office and national market surveillance authorities began enforcing Article 50 on August 2, gaining authority to request system access, demand documentation, order products withdrawn from the EU market, and impose fines. Eight foundation models have been identified as exceeding the systemic-risk training threshold and now must submit monthly evaluations.

China’s Interim Measures for AI Anthropomorphic Interactive Services took effect July 15 and produced penalties quickly. By the end of its first week, 12 companies had received fines totaling 4.2 million RMB, a pace contrasted with the absence of US fines under AI-specific rules.

The US federal picture remains less clear. A June executive order required a model evaluation framework within 60 days, but a meeting with Meta, Nvidia, Microsoft, OpenAI, Anthropic, and others took place on August 4, three days past the deadline. The framework will not be published, participation by AI labs is voluntary, and a Council on Foreign Relations senior fellow described the approach as “baffling.”

Other developments include the Pacing the Frontier statement from 1,238 AI lab employees, including Dario Amodei, Jakub Pachocki, and Shane Legg, calling for international coordination mechanisms; the UK AI Safety Bill clearing the House of Commons with Royal Assent expected in October; WAICO’s 29-country founding without Western democracies; the Take It Down Act taking effect as the first US federal synthetic media law; and active state-level efforts in Colorado, California, New York, Illinois, and Texas.