OpenAI Releases Proposal for International AI Safety Coordination

OpenAI released a proposal on Monday that would create international coordination around AI safety. In a blog post, OpenAI called for national AI safety institutes, such as the U.S. Commerce Department’s Center for AI Standards and Innovation, to set standards around areas including model evaluation, risk assessment and incident reporting. The proposal said that the standards would not serve as a system of licenses or mandatory prelease review, which has been a sticking point in U.S. negotiations for safety measures.

OpenAI has previously advocated for international action. In a blog post earlier this month, global affairs head Chris Lehane argued that the U.S. needed to first establish standards at home in order to lead internationally, though he added that international standards would soon be necessary. Any U.S. action, however, has hit stumbling blocks, as the White House waffles on regulatory policy and Congress debates a wide-range of legislative options. Meanwhile, OpenAI, Google, and Anthropic continue to discuss an industry-led standards body, The Information previously reported.

OpenAI’s proposal on Monday argued that international AI standards will be crucial as AI research moves toward recursive self-improvement, or the concept of AI systems being able to train themselves. Those standards should include human oversight over automated AI research and reporting around alignment, OpenAI said.

The post comes ahead of a summit between the U.S. and China, where the two superpowers will discuss AI development and safety. Treasury Secretary Scott Bessent met a Chinese delegation on Sunday night in Manhattan. They discussed a possible notification mechanism for AI incidents that rise to the national security level, Bessent told reporters.