Ahead of Sam Altman's UN address, OpenAI proposes new ways to track AI misalignment risks

OpenAI has unveiled a set of international AI safety standards as global leaders, including the United States and China, grapple with ways to reduce AI risks. The U.S. strategy for collaborating on AI risk with China will be significantly influenced by industry advice. Key developments: U.S. Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng recently talked about a U.S.-China AI dialogue to alert the other nation when AI incidents reach national security risk levels.

The specifics remain to be determined, but OpenAI's CEO, Sam Altman, will address these issues at the United Nations Security Council in New York on Wednesday. OpenAI suggests utilizing AI safety institutes worldwide to help establish standards through the U.S. Center for AI Security and Innovation, which falls under the Commerce Department.

The corporation also advocates for shared global metrics for categorizing incidents and how to monitor, report, and react to alignment problems. The aim is to have international standards for incidents that occur during testing in the pre-deployment phase and those that arise in the real world, according to an OpenAI official speaking with Axios.

Moving forward, anticipate AI firms continuing to contribute as President Trump and Chinese President Xi Jinping meet, and a state dinner featuring industry leaders takes place later in the week.