Trump signs order for voluntary pre-release AI model reviews
New executive order lets AI companies share models 30 days early, but it's optional.
President Donald Trump signed an executive order Tuesday establishing a voluntary framework for AI companies to share their most advanced (frontier) models with the federal government before public release. The order tasks several agencies with developing a process to “assess the advanced cyber capabilities of AI models” and prepare cyber defenses for critical infrastructure. Companies can choose whether to participate, and those that do will receive confidentiality protections. The timeline requires sharing models up to 30 days before launch, a reduction from an earlier draft that proposed 14 to 90 days.
This marks a notable shift for the Trump administration, which previously downplayed AI safety concerns and took a hands-off approach under former AI czar David Sacks. The change follows Anthropic’s limited April rollout of its Mythos model, which the company said flagged thousands of high-severity vulnerabilities across major operating systems and browsers. Google, Microsoft, and xAI agreed last month to pre-release reviews by the Commerce Department’s Center for AI Standards and Innovation (CAISI). OpenAI and Anthropic already participated in similar reviews under Biden. The order has drawn praise from groups like Americans for Responsible Innovation, though critics note it remains voluntary and lacks mandatory requirements.
- Voluntary framework allows AI companies to share models up to 30 days before release for cyber capability assessment.
- Google, Microsoft, and xAI have already agreed to participate; OpenAI and Anthropic had earlier agreements under Biden.
- Anthropic's Mythos model, which flagged thousands of vulnerabilities, helped prompt the administration's shift toward oversight.
Why It Matters
Signals growing government interest in AI security, but voluntary nature may limit real impact on risk mitigation.