US agencies could review frontier AI models for 30 days before public release

kimmonismus · x · 2026-07-28

The US is reportedly close to finalizing a framework that would let federal agencies review frontier AI models for up to 30 days before they are released.

According to the post, OpenAI, Anthropic, and Google are already negotiating the rules with the government. The NSA and CAISI would examine models for national-security risks, especially advanced cybersecurity capabilities. One unresolved issue is how to define a “frontier model,” including whether open-weight and closed-weight systems should be treated differently.

Related event: Trump Administration Nears Voluntary AI Regulatory Framework with Pre-Release Reviews(6 posts)→

Original post →

More from Safety

Safety channel →