Trump's AI order cuts review to 30 days, keeps framework voluntary

President Trump signed an executive order June 2 that replaces a withdrawn draft which "could've been a blocker" to US artificial intelligence development, adopting a voluntary 30-day government review framework that explicitly prohibits mandatory licensing.

"The signed order points toward some elements of a workable policy," said Steven E. Koonin, a senior fellow at Stanford University's Hoover Institution and former undersecretary of energy for science. "Voluntary sharing of frontier models with government evaluators creates accountability without pretending to control global development."

The order directs the Treasury Department, the National Security Agency and the Cybersecurity and Infrastructure Security Agency to develop a classified benchmarking process for assessing advanced AI models within 60 days. It also establishes an AI cybersecurity clearinghouse within 30 days and prioritizes enforcement against criminal actors using AI for unauthorized system access. The framework gives the federal government up to 30 days of early access to "covered frontier models" before they reach other trusted partners — down from the 90 days in the original draft.

The shift from mandatory to voluntary oversight removes a regulatory overhang that threatened to slow US labs relative to Chinese competitors. But Koonin argued the order does not confront the core problem: "Increasingly capable AI will be developed globally regardless of what Washington decides," with state-sponsored programs and open-source researchers entirely beyond the reach of any US executive order.

The order, titled "Promoting Advanced Artificial Intelligence Innovation and Security," marks the administration's most significant AI policy action since January 2025, when Trump revoked prior AI policies through Executive Order 14179. The White House had scrapped an earlier draft two weeks ago following industry pushback, with White House AI advisor David Sacks leading the charge against language that would have required 90 days of government review.

The voluntary structure vindicates the spending plans that major AI developers had already committed to. Nvidia reported first-quarter revenue of $81.6 billion, up 85% year over year, with its data center segment contributing $75.3 billion. Microsoft disclosed an AI business at a $37 billion annual run rate, up 123% year over year. Alphabet guided 2026 capital expenditure to between $175 billion and $185 billion, while Meta raised its 2026 capex guidance to between $125 billion and $145 billion.

The trigger for the administration's heightened security focus was Anthropic's unreleased Claude Mythos model, which autonomously identified thousands of previously unrecognized high-severity zero-day vulnerabilities in major operating systems, according to reports. The Center for AI Standards and Innovation, formerly the US AI Safety Institute, has been ordered to stop publishing its model evaluations, shifting assessments to a classified framework run by national security agencies.

The order's explicit disclaimer — that "nothing shall be construed to authorize the creation of a mandatory governmental licensing, preclearance or permitting requirement" — is significant for developers concerned about regulatory drag. OpenAI has already diverged from the administration's approach, releasing a policy paper calling for mandatory federal evaluations of advanced AI models overseen by civilian agencies rather than national security bodies.

For critical infrastructure operators including rural hospitals, community banks and local utilities, the order promises expanded access to AI-enabled cybersecurity tools through new Binding Operational Directives from the Department of Homeland Security. The planned AI cybersecurity clearinghouse will coordinate vulnerability scanning, validation and patch distribution in voluntary collaboration with the AI industry.

The last time the administration issued a major AI directive was December 2025, when Executive Order 14365 established an AI Litigation Task Force to challenge state-level AI laws. The June 2026 order represents a shift from that anti-regulatory posture toward affirmative national security engagement, though it maintains the voluntary, industry-friendly framework that developers sought.

This article is for informational purposes only and does not constitute investment advice.