The United States is developing a national AI safety framework through a process of 'reverse federalism,' where individual state actions catalyze broader federal standards. California, New York, and Illinois have already moved to pass frontier safety legislation, creating a de facto national baseline. These state-level efforts focus on three primary requirements: establishing a documented safety framework with public disclosure of risk assessments, mandated reporting of serious safety incidents, and ensuring governance through independent, objective audits. California initiated the disclosure framework, New York demonstrated jurisdictional scalability, and Illinois added the requirement for independent verification of disclosures.

At the federal level, the Trump Administration is working with national security and technical experts to establish a framework for government-led testing of the most capable AI models. This initiative aims to define clear testing standards and timelines to facilitate the secure deployment of advanced AI tools to government agencies, critical infrastructure defenders, and international allies. OpenAI recommends that this federal work strengthen the Center for AI Standards and Innovation (CAISI), implement clear requirements for companies such as independent audits and whistleblower protections, and ensure that federal policies reinforce state-level efforts.

These combined national strategies serve as the foundation for a US-led global AI framework. Recent discussions at the G7, involving partners including Brazil, Egypt, India, Kenya, and Korea, have underscored the push for international standard-setting. Following these talks, OpenAI CEO Sam Altman proposed an international forum to establish accepted standards, provide impartial analysis of AI capabilities, and manage technology access for participating nations. This multi-layered approach aims to ground the global development of AI in a democratic vision, ensuring that essential governance decisions are made by elected governments rather than solely by private frontier labs.

The United States is developing a national AI safety framework through a process of 'reverse federalism,' where individual state actions catalyze broader federal standards. California, New York, and Illinois have already moved to pass frontier safety legislation, creating a de facto national baseline. These state-level efforts focus on three primary requirements: establishing a documented safety framework with public disclosure of risk assessments, mandated reporting of serious safety incidents, and ensuring governance through independent, objective audits. California initiated the disclosure framework, New York demonstrated jurisdictional scalability, and Illinois added the requirement for independent verification of disclosures.

At the federal level, the Trump Administration is working with national security and technical experts to establish a framework for government-led testing of the most capable AI models. This initiative aims to define clear testing standards and timelines to facilitate the secure deployment of advanced AI tools to government agencies, critical infrastructure defenders, and international allies. OpenAI recommends that this federal work strengthen the Center for AI Standards and Innovation (CAISI), implement clear requirements for companies such as independent audits and whistleblower protections, and ensure that federal policies reinforce state-level efforts.

These combined national strategies serve as the foundation for a US-led global AI framework. Recent discussions at the G7, involving partners including Brazil, Egypt, India, Kenya, and Korea, have underscored the push for international standard-setting. Following these talks, OpenAI CEO Sam Altman proposed an international forum to establish accepted standards, provide impartial analysis of AI capabilities, and manage technology access for participating nations. This multi-layered approach aims to ground the global development of AI in a democratic vision, ensuring that essential governance decisions are made by elected governments rather than solely by private frontier labs.