June, Only Apply to Closed-source AI Models With State-of-the-art Capabilities and National Security Risks, But the Framework Doesn't Define

The Trump administration's framework for assessing cybersecurity risks from advanced AI excludes open models entirely, according to Axios. The voluntary guidelines, created after President Trump signed an executive order in June, only apply to closed-source AI models with state-of-the-art capabilities and national security risks, but the framework doesn't define those terms. It sets a 30-day review period and can't restrict open models after release.

AI companies including Anthropic, OpenAI, and Google attended a White House briefing on the finalized framework, but details won't be made public. Compliance is voluntary, though frontier labs seek guidance to avoid government restrictions, leaving smaller providers uncertain.