President Donald Trump signed an executive order on June 2 that asks technology companies to provide the federal government with a preview of their advanced artificial intelligence (AI) models before releasing them to the public.
The order represents a narrower approach than a draft version that the administration reportedly shelved more than a week ago. It focuses on voluntary pre-release collaboration between AI developers and the government.
The White House said the effort is intended to address national security concerns tied to increasingly capable frontier AI systems while maintaining close coordination with industry.
“Advanced AI capabilities make our Nation stronger, but also introduce new national security considerations that require coordinated action across executive departments and agencies (agencies), and components,” the order says.
“As these capabilities evolve, my Administration will continue to work closely with industry to ensure that the best and most secure technology is deployed rapidly to confront any and all threats to our country,” it adds.
Under the executive order, the administration will create a voluntary framework through which developers of advanced AI models can provide the federal government an opportunity to evaluate certain frontier systems before they are publicly released.
It notes that the voluntary testing would allow the federal government access to frontier AI models “up to 30 days before they plan to release such models to other trusted partners.”
The order also tasks several agencies – including the Department of Defense, Department of the Treasury, and Cybersecurity and Infrastructure Security Agency – with strengthening U.S. cyber defenses to address emerging threats posed by advanced AI capabilities.
The directive arrives amid growing concern in Washington about the cybersecurity implications of frontier AI models.
Policymakers and federal security officials have increasingly cited Anthropic’s Claude Mythos Preview model – which the company unveiled in April – as evidence that advanced AI systems are changing the cyber threat landscape rapidly. Mythos, for example, has powerful, native hacking capabilities.
Coinciding with the executive order, Anthropic announced today that it is expanding its Project Glasswing partnership to approximately 150 new organizations, giving them early access to Claude Mythos Preview. The company said the expansion comes after “several weeks of close collaboration” with the U.S. government and other Project Glasswing partners.
Similarly, last month, OpenAI announced a new tool – labeled GPT-5.5-Cyber – that has been touted as a competitor to Mythos.
The executive order does not establish mandatory testing requirements or pre-release approvals, instead relying on voluntary participation by AI developers as the administration evaluates additional policy options.