Why It Matters
Artificial intelligence is increasingly viewed as a front line in the competition between the United States and China for global technological dominance. Tuesday’s executive order represents the Trump administration’s attempt to address legitimate national security concerns about advanced AI without placing restrictions that could slow American innovation — a balance the President himself has emphasized publicly.
“We’re leading China, we’re leading everybody, and I don’t want to do anything that’s going to get in the way of that lead,” Trump said in remarks tied to the signing.
What Happened
President Trump on Tuesday signed an executive order creating a federal structure to evaluate advanced AI systems for potential national security risks ahead of public deployment. The order assigns the Director of the National Security Agency to lead the government’s review effort, giving federal agencies a 30-day window to assess any AI system submitted for evaluation.
Critically, the program is entirely voluntary — developers are not required to submit their models, and the White House was explicit that the order does not extend government oversight to every new AI product brought to market. The framework applies only to frontier AI laboratories that choose to share their most advanced systems with the government.
The signing came less than two weeks after Trump called off a scheduled May 21 Oval Office ceremony with technology company executives. At that time, Trump declined to put his signature on an earlier draft of the policy, with the administration citing concerns that the directive as written could blunt the competitive edge American AI companies hold over foreign rivals. A revised version ultimately satisfied those concerns.
Industry Response
The three largest U.S. AI developers — Anthropic, OpenAI, and Google — each expressed support for the new framework. OpenAI’s chief global affairs officer, Chris Lehane, said the company views collaborative, institutionally grounded safety measures as the right approach. “As AI capabilities continue to advance, we believe effective safety frameworks should continue to be developed through democratic institutions, informed by technical expertise and broad stakeholder input, to promote accountability and public trust,” Lehane said in remarks reported by major outlets.
Anthropic has been at the center of recent discussions over AI risk. The company revealed a powerful new model called Claude Mythos in April — a disclosure that prompted Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell to call an urgent meeting with Wall Street chief executives to assess the dangers posed by Mythos’ reported ability to detect cybersecurity weaknesses in software systems. Anthropic initially kept Mythos access tightly restricted, but expanded its circle of trusted partners by 150 organizations as of Tuesday. The company’s Mythos announcement also came while it is engaged in a legal dispute with the Trump administration related to a Pentagon contract.
By the Numbers
- 30 days: the review period federal agencies have to evaluate an AI system submitted under the new framework
- 150 organizations: the number of new partners Anthropic added to its trusted access group for Claude Mythos
- May 21: the date of the canceled Oval Office signing ceremony and Trump’s initial refusal to sign the earlier order
- Under 2 weeks: the time elapsed between the canceled ceremony and Tuesday’s completed signing
Zoom Out
The new order stands in stark contrast to the regulatory posture of the prior administration. Within hours of taking office in January 2025, Trump repealed the AI oversight rules President Biden had put in place. Democratic Sen. Mark Warner criticized the administration for eliminating those guardrails during its first year and taking time to establish any alternative structure in their place.
The administration has consistently argued that heavy-handed regulation of AI poses a greater threat to U.S. leadership than the risks it purports to address. The voluntary design of Tuesday’s order reflects that philosophy — prioritizing industry cooperation over government mandates while still creating a channel for security-sensitive AI systems to be evaluated before they reach the broader market.
What’s Next
Federal agencies will begin building out the review infrastructure called for in the order. Companies that opt into the program will have their submissions assessed within the 30-day timeline. Whether the voluntary structure proves sufficient to address the most serious AI security concerns will depend heavily on how many developers choose to participate — and OpenAI, Google, and Anthropic’s public endorsements suggest at least the largest players intend to engage with the process.
About the Author
