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

OpenAI released a set of international AI safety standards on Monday, according to Axios. The company recommends leveraging AI safety institutes around the world to facilitate standard-setting, and the proposal routes that work through the U.S. Center for AI Security and Innovation, which the article identifies as part of the Commerce Department.

The same report states that U.S. Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng discussed a U.S.-China AI dialogue in which the two countries would notify each other when AI incidents reached national security risk levels. According to the report, the proposal comes as those two officials held that discussion. Axios presents the OpenAI proposal and the bilateral discussion as concurrent developments; the article's headline describes the proposal as new ways to track AI misalignment risks ahead of Altman's UN address.

Axios does not list the individual standards OpenAI is proposing, nor does it name any safety institutes beyond the U.S. center that the company suggests involving. The report does not state whether any government has endorsed the proposal, whether the standards are binding or voluntary, or whether the U.S.-China dialogue discussed by Bessent and He has been formally established. The article also does not describe what threshold or process would define an AI incident reaching national security risk levels.