The United States has rejected calls from leading AI companies for global standards on evaluating AI risks, pushing back against pleas from OpenAI’s Sam Altman, Anthropic’s Dario Amodei and Hugging Face’s Clement Delangue for internationally coordinated safety frameworks.

The rebuff highlights the growing tension between the AI industry’s warnings about the technology’s risks and governments’ reluctance to bind themselves to global rules.

What the companies asked for

The AI leaders have argued that frontier AI systems are advancing so fast that no single country can manage their risks alone. They have called for shared standards on risk evaluation, testing and deployment, similar in spirit to international regimes that govern nuclear or aviation safety.

Altman, Amodei and Delangue have each warned that without coordination, a race to the bottom could see safety sacrificed for speed.

Why Washington said no

The US rejection reflects a preference for national approaches to AI governance and a wariness of international frameworks that could constrain American companies or hand leverage to geopolitical rivals. Washington has instead pursued its own regulatory and voluntary-commitment tracks.

Critics of the decision argue that AI risks do not respect borders and that fragmented national rules will leave dangerous gaps.

What happens now

The debate over AI governance is far from settled. With the technology advancing rapidly and its societal impact widening, pressure for some form of international coordination is likely to grow.

Whether that pressure eventually produces binding global standards, or whether national regimes remain the norm, will be one of the defining policy questions of the AI era.

Other countries and blocs, including the European Union, have moved ahead with their own AI rulebooks, creating a patchwork of approaches that companies must navigate.

For now, the industry’s biggest players find themselves in the unusual position of asking for more regulation while governments hesitate to impose it.