EU, US and China split over AI rules

Pressure is rising worldwide to control the most powerful AI systems, with shared concerns over AI-powered cyberattacks, national security risks and chatbots linked to children’s suicides. The EU, US and China are responding with sharply different regulatory models shaped by economic priorities and geopolitical competition.

The EU adopted its AI Act in 2024, setting binding rules based on levels of risk and banning systems such as social scoring, emotion recognition and manipulation. Industry pressure has already led to delays for high-risk provisions under the AI Omnibus bill, while Brussels is also using the Digital Services Act and a proposed EU Kids Act to scrutinize chatbots and AI companions, including systems that may create emotional dependency in children.

The US has favored voluntary commitments and executive orders under President Donald Trump. Trump revoked Joe Biden’s 2023 order requiring major AI developers to share safety test results with the government, and in July 2025 the White House released an AI Action Plan with more than 90 measures to accelerate development, infrastructure investment and deregulation. A later order created voluntary early access of up to 30 days for powerful models, while on September 29 heads of America’s six largest AI companies signed a self-policing accord with no penalties.

China is building detailed rules under tight state control. Companion chatbot regulations took effect in July, and President Xi Jinping said AI should be “always under human control.” Beijing continues to reject a broad slowdown, while censorship rules and national security warnings frame generative AI as both a strategic priority and a risk.