The Trump administration is considering an executive order aimed at ensuring the security of artificial intelligence models prior to their public release, as stated by Kevin Hassett, director of the National Economic Council. This potential policy shift comes in response to rising concerns about cybersecurity risks associated with AI, particularly following Anthropic’s revelation of its advanced “Mythos” model, which can exploit software vulnerabilities. The administration’s move toward greater AI oversight signals a departure from its earlier hands-off approach and may increase the responsibilities of the Center for AI Standards and Innovation (CAISI), which has recently established partnerships with tech companies like Google DeepMind and Microsoft for pre-deployment evaluations of AI models. However, there are concerns regarding CAISI’s limited resources, as experts indicate it is underfunded to effectively carry out its expanding mandate for AI safety evaluations.

Chris Fall: Chris Fall was recently appointed director of the Center for AI Standards and Innovation (CAISI), drawing on prior experience as an Energy Department official. Under his leadership, CAISI has forged new pacts with major AI developers for rigorous pre-deployment evaluations of frontier models. He has stressed the importance of independent measurement science in industry collaborations to address national security implications of AI.
Kevin Hassett: Kevin Hassett directs the White House National Economic Council, providing economic policy advice that now extends to AI governance and security. In a recent Fox Business interview, he outlined the administration’s consideration of an executive order to create a roadmap for vetting advanced AI models for vulnerabilities before release. His comments directly addressed risks highlighted by Anthropic’s Mythos model to American businesses and government.
Sean Cairncross: Sean Cairncross holds the position of National Cyber Director, leading coordination of cybersecurity initiatives across government and private sectors. He is overseeing an all-of-government response to test powerful AI models like Anthropic’s Mythos for potential harms prior to wider release. White House officials have expressed high confidence in his efforts to mitigate cyber threats posed by such technologies.
Trump Administration: The Trump Administration leads the U.S. federal executive branch and has recently pivoted from a largely hands-off stance on AI toward proposing greater oversight to safeguard national security. Officials are studying an executive order that would require new AI models to undergo a pre-release safety process akin to FDA drug evaluations before public deployment. This shift was prompted by Anthropic’s Mythos model, which demonstrated rapid vulnerability discovery capabilities.
Center for AI Standards and Innovation: The Center for AI Standards and Innovation (CAISI) operates within the Commerce Department’s National Institute of Standards and Technology to establish standards and perform evaluations on frontier AI systems. Rebranded under the Trump administration from the prior AI Safety Institute, it serves as the primary government hub for AI testing, research, and best practices with industry. Recently, CAISI expanded agreements with Google DeepMind, Microsoft, and xAI for pre-deployment national security assessments of their models.

Policy Shift: The Trump administration is embracing AI oversight measures it previously rejected, spurred by cybersecurity concerns from Anthropic’s Mythos model.
Resource Concerns: Experts and think tanks have flagged CAISI as under-resourced for its growing responsibilities in AI safety evaluations.
Industry Collaborations: CAISI announced recent agreements with Google DeepMind, Microsoft, and xAI to enable pre-deployment testing of frontier AI models for national security risks.