OpenAI Proposes Global AI Safety Standards Ahead of UN Address

OpenAI released a framework for international AI safety standards Monday, positioning itself at the center of emerging discussions between the United States and China on coordinating responses to AI-related national security risks.

The timing coincides with weekend talks between U.S. Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng, who discussed establishing a bilateral AI dialogue focused on notifying each government when AI incidents reach national security thresholds. OpenAI CEO Sam Altman is scheduled to present the company's recommendations to the United Nations Security Council on Wednesday.

Why it matters

How the U.S. government approaches AI risk coordination with China will be shaped significantly by recommendations from major AI companies. OpenAI's proposal offers a concrete blueprint for international cooperation at a moment when both superpowers are seeking mechanisms to manage AI deployment without stifling development.

The framework details

OpenAI's standards proposal centers on leveraging existing AI safety institutes worldwide to establish unified protocols. The company recommends routing standard-setting efforts through the U.S. Center for AI Safety and Innovation, housed within the Commerce Department.

The framework calls for shared global measurements that would classify AI incidents consistently across borders. This includes standardized methods for tracking, reporting, and responding to alignment issues—situations where AI systems behave in ways that diverge from intended parameters.

According to an OpenAI official who spoke with Axios, the standards would apply to two distinct categories: incidents detected during pre-deployment testing and those occurring after systems are released into real-world use. This dual approach acknowledges that risks manifest differently in controlled laboratory environments versus production settings.

Political context

The release comes as President Trump and Chinese President Xi Jinping prepare for meetings and a state dinner later this week that will include prominent figures from the AI industry. The convergence of diplomatic engagement and industry input suggests AI governance is moving from theoretical discussion to practical implementation.

Industry observers expect other major AI laboratories to issue their own recommendations as the U.S.-China dialogue takes shape. The details of how incidents would be classified, what thresholds would trigger notifications, and how information would be shared between governments remain to be defined.

The proposal reflects OpenAI's positioning as both a commercial AI developer and a participant in shaping the regulatory environment that will govern its operations. By advocating for international standards before governments impose unilateral requirements, the company seeks influence over the framework that will ultimately constrain or enable its business model.

These details were first reported by Axios.