President Donald Trump signed an executive order on June 2, 2026, laying the foundation for federal testing of the world's most powerful AI systems before they reach the public. The order asks leading AI companies to voluntarily hand over their frontier models so the government can assess them for up to 30 days ahead of release.
Crucially, the order explicitly bars any mandatory licensing or pre-clearance requirement, making the review a request rather than a rule. It relies on voluntary cooperation from companies like Anthropic, OpenAI and Google, and also directs federal agencies to build benchmarks for AI cyber capabilities and to create an "AI cybersecurity clearinghouse" to share information on vulnerabilities.
The move marks a sharp shift from the administration's earlier hands-off stance. An initial version of the order, which would have allowed up to 90 days for review, was scrapped last month over fears it would stifle innovation before being trimmed to 30 days. Reporting suggests the push was partly prompted by Anthropic's Mythos model, which the company itself judged too risky to release publicly.