The U.S. pressed Meta to submit frontier AI models for pre-release review

According to a New York Times report carried by Reuters, the U.S. government is pressing Meta to sign a voluntary agreement allowing federal review of certain frontier AI models before release or partner access. The review would assess capabilities, vulnerabilities, and national-security risks. OpenAI, Anthropic, and others have already participated in related testing. Meta said it shares the goal of improving frontier AI safety while details remain under discussion.