President Trump signed an executive order creating a voluntary program for AI firms to submit advanced models for a 30-day government security review before public release. The order, which follows a postponement due to tech industry lobbying, aims to address national security risks from AI like Anthropic's Mythos without imposing mandatory regulations that could hinder the U.S. lead over China.
Voluntary AI model vetting framework
- ▪Under the program, AI developers are asked to provide the government with access to their models for up to 30 days before public release
- ▪The order directs the departments of Treasury, Defense, Commerce, and Homeland Security to secure voluntary testing agreements with AI developers
- ▪The executive order explicitly states it does not authorize mandatory government licensing, preclearance, or permitting for new AI models
- ▪President Donald Trump signed an executive order on June 2, 2026, creating a voluntary program for AI companies to submit new models for government review
Tech industry lobbying influence
- ▪The postponement of an earlier, stricter version of the executive order was reportedly influenced by lobbying from Silicon Valley leaders
- ▪Tech billionaires including Elon Musk, Mark Zuckerberg, and David Sacks privately lobbied President Trump against a more restrictive AI order
Executive order postponement
- ▪An earlier draft of the order reportedly requested access to AI models 90 days before their release, compared to the final 30-day period
- ▪President Trump postponed signing an earlier version of the AI order on May 21, 2026
- ▪Trump stated he delayed the signing because he feared it would hinder the U.S. competitive lead over China in AI
National security threats
- ▪The National Security Agency and the Department of Defense will help identify which AI models require government scrutiny under the new framework
- ▪Anthropic's Mythos model has raised security concerns due to its advanced capability to find and exploit software vulnerabilities at an unprecedented scale
- ▪The executive order establishes a framework to vet the national security risks of the most advanced AI systems
Cybersecurity vulnerabilities
- ▪The order directs Treasury Secretary Scott Bessent to work with AI firms and critical infrastructure providers to scan for and patch software vulnerabilities
- ▪The order calls for establishing an AI cybersecurity clearance house involving multiple secretaries and AI operators to find software flaws
- ▪The order directs the government to hire more cybersecurity and AI professionals and strengthen cyber defenses for critical infrastructure like hospitals and banks
Existing federal AI review agreements
- ▪Some companies, including OpenAI and Anthropic, already had voluntary agreements to submit models to the Commerce Department's Center for AI Standards and Innovation
- ▪Details of the May 2026 security testing agreement with Microsoft, Google, and xAI were later removed from a U.S. government website
- ▪In May 2026, the government announced that Google, xAI, and Microsoft had also agreed to submit their AI models for security testing
Story comments
Loading comments…