White House Considers Vetting A.I. Models Before They Are Released

The White House, under the Trump administration, is actively considering a policy to require pre-release review of advanced AI models before they are made public. The trigger for this shift is Anthropic's latest model, Mythos, which has raised national security concerns. However, the proposal currently includes no published criteria for what would trigger a review—no alignment benchmarks, safety thresholds, or capability metrics. This lack of transparency could turn the process into a discretionary lever, subject to political or contractual disputes.

This concern is not theoretical. The Pentagon recently cut off use of Anthropic's technology over a $200 million contract dispute, and Anthropic has since sued. Such friction could extend to any lab—large or small—that finds itself in conflict with the administration. The bigger risk is competitiveness: lead time in AI deployment accrues to those who ship without waiting for review. Today, that means Chinese labs, who face no equivalent barrier. Pre-release vetting, if poorly defined, could slow adoption in defense and security sectors and inadvertently cede the AI race to global rivals.