Trump Signs Order for Voluntary National Security Review of Advanced AI Models
Summary
Trump's order creates a voluntary federal process to review advanced AI systems for national-security risks for up to 30 days before release. The White House says frontier labs may share cyber models to support critical-infrastructure security and government defenses.
Key points
- The order allows federal agencies up to 30 days to review advanced AI systems for national-security risks before public release; developer participation is voluntary.
- The White House says frontier labs may voluntarily share cutting-edge cyber models to help secure critical infrastructure and strengthen government cyber defenses.
- The article says it was unclear how the signed order differed from an earlier version Trump declined to sign.
- Cato Institute analyst Juan Londoño raised concerns about the NSA director's discretion in deciding which models qualify for review and which partners receive early access.
- OpenAI called the policy an important step; Sen. Mark Warner welcomed it while criticizing the administration's earlier rollback of AI guardrails.
- The policy followed Anthropic's announcement of Claude Mythos, which the article says appeared able to find cybersecurity vulnerabilities; Anthropic said it had expanded access to another 150 organizations.
Article Details
- Event Type
- Executive order establishing voluntary federal review of advanced AI models for national security risks
- Impact
- Creates a process for participating AI developers to share advanced models for government review for up to 30 days before public release, with the stated aim of protecting critical infrastructure and strengthening cyber defenses. The article reports concerns that advanced models such as Claude Mythos may identify cybersecurity vulnerabilities in widely used software; the order’s criteria for model scrutiny and access by trusted partners were described as vague.
People
Chris LehaneOpenAI’s chief global affairs officer, whose statement described the policy as an important step toward developing AI safety frameworks through democratic institutions.Donald TrumpSigned the executive order establishing the voluntary review framework for advanced AI models.Jerome PowellOutgoing Federal Reserve Chair who, with Scott Bessent, convened a meeting with Wall Street CEOs and warned about risks attributed to Claude Mythos’s apparent ability to find software vulnerabilities.Joe BidenFormer president whose AI safeguards Trump repealed shortly after returning to the White House, as described in the article.Juan LondoñoCato Institute policy analyst who assessed the order as a step toward preparing for advanced AI systems and raised concerns about the NSA director’s discretion over model selection and early access.Mark WarnerDemocratic senator and vice chairman of the Senate Intelligence Committee who welcomed the policy while criticizing the administration’s earlier dismantling of AI safeguards.Scott BessentTreasury Secretary who, with Jerome Powell, convened a meeting with Wall Street CEOs and warned about risks attributed to Claude Mythos’s apparent ability to find software vulnerabilities.
Vendors
Anthropicwas characterized as a voluntary collaboration with participating U.S.-based tech companies, including Anthropic, OpenAI and Google, which are sometimes described as “frontier labs” because they are building the mostGooglea voluntary collaboration with participating U.S.-based tech companies, including Anthropic, OpenAI and Google, which are sometimes described as “frontier labs” because they are building the most advanced AI systems.OpenAIas a voluntary collaboration with participating U.S.-based tech companies, including Anthropic, OpenAI and Google, which are sometimes described as “frontier labs” because they are building the most advanced