Administration Issues Voluntary AI Model Review Order

The order establishes a voluntary framework under which AI companies may submit frontier models to federal agencies for review up to 30 days before public release. It directs companies to provide the federal government access to covered frontier models subject to confidentiality, cybersecurity, insider-risk, and intellectual-property protections.

The order gives the Pentagon, Department of Homeland Security, Cybersecurity and Infrastructure Security Agency, Office of Management and Budget, and other agencies 30 days to expedite cyber defense of civilian federal systems and to expand AI-enabled defensive tools.

It also creates an AI cybersecurity clearinghouse that will operate in voluntary collaboration with industry and critical infrastructure operators to scan frontier models for vulnerabilities and distribute patches.

April the Commerce Department's Center for AI Standards and Innovation announced partnerships with Google, Microsoft, and xAI to test their frontier models for security risks ahead of release. Those tests continue under the new framework. Anthropic announced its Claude Mythos Preview model on April 7.

The model remains unavailable to the public because the company said bad actors could use it to discover critical software exploits. The Trump administration previously moved to bar Anthropic from federal contracts after the company declined to grant the Pentagon unrestricted access to its Claude models.

Anthropic co-founder Jack Clark stated in April that the company had been in talks with the administration over the Mythos model.

The Alliance for Secure AI called on June 2 for Congress to enact legislation making federal review of advanced models mandatory. Brendan Steinhauser, CEO of the Alliance for Secure AI, said the next generation of models will be more powerful and pose greater threats, adding that companies need oversight and cannot be trusted to act voluntarily.

“The next generation of models will be more powerful and pose greater threats.”