A new proposal from OpenAI seeks a coordinated international effort to evaluate technical capabilities and assess risks in frontier models. This framework identifies the United States as the central authority for overseeing common incident reporting and human review protocols. The plan tracks recursive self improvement gains and autonomous research volume to prevent systems from becoming too opaque, though OpenAI specified that these standards should not become mandatory model licenses or government approval requirements.

These guidelines overlap with U.S. and China negotiations over a notification mechanism for AI incidents that reach the national security level. Treasury Secretary Bessent stated the two countries agreed to meet again under a U.S.-China AI dialogue to increase transparency between the top two AI powers. U.S. trade representative Jamieson Greer noted that chip export controls are not on the agenda for these talks.