The White House is reviewing the possibility of issuing an executive order that would establish a formal vetting process for newly developed artificial intelligence models, National Economic Council Director Kevin Hassett said on Wednesday.
Under the proposed framework, AI systems would be required to undergo testing prior to being released publicly - a process Hassett compared to the way the Food and Drug Administration evaluates new drugs. He described the idea during remarks to Fox Business.
"We have scrambled an all of government effort and all the private sector to coordinate and make sure that before this model is released out into the wild, that it's been tested left and right, to make sure that it doesn't cause any harm to the American businesses or the American government," Hassett said.
Hassett framed the initiative as an interagency and public-private coordination effort intended to identify risks and limit potential harm to businesses and government systems. He indicated that the vehicle for such coordination could be an executive order that mandates testing prior to public deployment of AI models.
The immediate impetus for the initiative includes Anthropic PBC's disclosure that its Mythos model can detect network vulnerabilities, a capability that the company said could pose cybersecurity risks. Anthropic has limited access to Mythos, allowing only select large technology and financial firms to use the model. Hassett confirmed that the Trump administration has pursued access to Mythos for federal agencies to test government systems.
White House Chief of Staff Susie Wiles and other senior administration officials held a meeting last month with Anthropic CEO Dario Amodei, during which Mythos was among the topics discussed, Hassett said.
Hassett suggested that required testing under any executive order would "really quite likely" extend to all AI firms. He characterized Mythos as the first high-profile example prompting consideration of wider safeguards, saying, "I think that, that Mythos is the first of them, but it's incumbent on us to build a system."
The precise contours and scope of any mandatory testing requirements remain unclear. Such measures would mark a change from President Donald Trump's prior emphasis on a lighter regulatory approach to artificial intelligence.
Separately, the Commerce Department recently expanded a voluntary AI testing program. Under that program - which is managed by the department's Center for AI Standards and Innovation - Alphabet Inc.'s Google, Microsoft Corp. and xAI agreed to let the government assess their models for capabilities and security improvements. OpenAI and Anthropic were already participants in the program.
How broadly any formal testing mandate would be applied, what standards would govern evaluations, and how enforcement would be structured are outstanding questions that the administration has not yet resolved, based on Hassett's statements.