AI rules recalibrate as targeted laws expand

Global AI regulation is moving through a recalibration rather than a broad retreat. The EU has simplified its AI Act and delayed some high-risk requirements by up to two years, while US executive action has pushed for a “minimally burdensome national policy framework” and challenged conflicting state rules. The UK has again delayed its AI bill, Canada lacks a replacement for its defunct federal proposal, and global AI safety summits are drawing less attention.

That pullback is being offset by more targeted activity. US states have accelerated AI rulemaking, with Texas, California, New York, Connecticut and Illinois among those adopting safety or transparency measures, and a wave of companion chatbot laws spreading from Hawaii to Rhode Island. There are now well over 250 US state AI laws covering areas including employment and deepfakes, while the federal government has used export controls to regulate access to some frontier models.

Outside the US, AI-specific rules are also advancing. Vietnam is rolling out its AI Law, Taiwan’s and South Korea’s new laws have taken effect, China has issued interim measures on AI companionship services, and India, Mexico, Brazil and Thailand are progressing reforms. General legal frameworks, including consumer, intellectual property, data, digital and product safety laws, continue to carry much of the regulatory burden, making jurisdiction-by-jurisdiction compliance mapping increasingly important.