EU, US and China split over AI rules

AI governance is diverging sharply across the EU, US and China. The EU AI Act moves into high-risk enforcement on August 2, 2026, with penalties reaching €35 million or 7% of global revenue for prohibited practices. Its risk-based framework is already shaping global compliance behavior, though Europe remains weak in frontier model production.

The US is prioritizing speed, capital and infrastructure after rolling back federal safeguards and challenging state-level AI rules. It still leads in private investment, with $285.9 billion in 2025, and has a major compute advantage at 39.7 million petaflops. But public trust is low, and its model lead over China narrowed from 9.26 percentage points in January 2024 to 2.7 percentage points by March 2026.

China is advancing through a layered regulatory stack, aggressive deployment and patent volume. It filed 38,210 generative AI patents between 2014 and 2023, compared with 6,276 for the US, while using licensing and service suspension as enforcement tools. Its biggest constraint is compute, with 400,000 petaflops, leaving no clear overall winner across innovation, standards and deployment.