The Trump administration drafts requirements for Google DeepMind, Microsoft, and xAI to share models with the Commerce Department's Center for AI Standards and Innovation (renamed from US AI Safety Institute in June 2025) for pre-release security testing. This follows Donald Trump's January 20, 2025 revocation of Biden's Executive Order 14110. No legal authority exists to require pre-release reviews. Anthropic's Mythos model found thousands of high-severity vulnerabilities.
Pre-release vetting discussions
- ▪The Trump administration discussions about AI model vetting center on systems capable of facilitating cyberattacks, particularly models that could help users identify and exploit software vulnerabilities
- ▪No proposal for AI model pre-release vetting has been finalized by the Trump administration, and no timeline has been set
- ▪Trump administration officials are considering options including formal pre-release review processes and government-led testing for higher-risk AI systems
- ▪The Trump administration is in early discussions about whether advanced AI models should be vetted before public release
Trump administration policy shift
- ▪Donald Trump revoked Biden's Executive Order 14110 on Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence on January 20, 2025
- ▪The article was published on May 5, 2026
- ▪Biden's Executive Order 14110 had required developers of the largest AI systems to notify the government and share safety test results before deployment
- ▪The Biden AI framework had introduced mandatory red-teaming for high-risk AI models, enhanced cybersecurity protocols, and monitoring requirements for AI used in critical infrastructure
- ▪Donald Trump issued an executive order titled "Removing Barriers to American Leadership in Artificial Intelligence" on January 23, 2025
Anthropic Mythos model
- ▪Anthropic stated that Mythos Preview has found thousands of high-severity vulnerabilities, including some in every major operating system and web browser
- ▪Anthropic launched Project Glasswing, committing up to $100 million in usage credits to a select group of technology and cybersecurity companies to use Mythos for defensive purposes
- ▪Anthropic has not released the Mythos model publicly
- ▪Anthropic introduced Mythos, an AI model the company has described as representing a watershed moment for cybersecurity
- ▪OpenAI has developed a model comparable to Anthropic's Mythos and has released it to a small set of companies through an existing trusted-access program
- ▪Anthropic has been briefing the Cybersecurity and Infrastructure Security Agency, the Commerce Department, and other stakeholders on the potential risks and benefits of Mythos Preview
US institutional changes
- ▪The US has not established a legal authority to require pre-release reviews of AI models
- ▪The US AI Safety Institute was renamed the Center for AI Standards and Innovation in June 2025
- ▪The Center for AI Standards and Innovation's mandate includes leading unclassified evaluations of AI capabilities that may pose risks to national security, with a stated focus on demonstrable risks such as cybersecurity, biosecurity, and chemical weapons
International AI governance
- ▪The UK's AI Security Institute has conducted pre-deployment evaluations of several frontier models, working directly with labs including Anthropic and OpenAI
- ▪The EU AI Act began phasing in in 2025 and establishes mandatory conformity assessments for high-risk AI applications
Perspective of Anthropic
- ▪Anthropic has been briefing the Cybersecurity and Infrastructure Security Agency, the Commerce Department, and other stakeholders on the potential risks and benefits of Mythos Preview
Story comments
Loading comments…