OpenAI calls for aligned US AI safety framework

OpenAI has called for closer alignment between US state and federal AI safety efforts, arguing that a common framework is needed to govern frontier AI systems.

In a policy blog post, the company said recent frontier AI legislation in California, New York and Illinois shows how states can help create a shared baseline before a single federal framework is in place.

OpenAI describes this process as ‘reverse federalism’, where state laws move in similar directions and gradually shape a de facto national standard.

The company says core elements should include documented safety frameworks, risk assessments for frontier models, public disclosure of results, serious incident reporting and independent audits.

At the federal level, OpenAI argues that the US government should lead testing and evaluation of the most advanced AI systems, particularly when national security and cybersecurity are at stake.

It says a consistent federal testing framework would help advanced AI tools reach trusted users, including government agencies, critical infrastructure defenders, allies and other partners.

OpenAI also supports clearer requirements for companies developing the most capable systems, including strong security standards, incident reporting, independent audits and whistle-blower protections.

The company warns that neither a fragmented patchwork of state laws nor an undefined federal process would create a coherent frontier safety regime.

OpenAI links domestic AI governance to global competition, arguing that a national US framework could support international AI standards built around democratic values.

Why does it matter?

OpenAI’s proposal highlights the growing tension in US AI governance between state-led action, federal oversight and international standard-setting. A shared framework could reduce regulatory fragmentation and create clearer expectations for frontier model developers. Still, the company’s position also reflects the interests of a major AI lab seeking predictable rules for deployment, testing and access. The debate will shape how the US balances safety, innovation, national security and global influence in AI governance.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!