President Trump signed the “Promoting Advanced Artificial Intelligence Innovation and Security” executive order in a private June 2 ceremony, twelve days after pulling a broader version that would’ve given the federal government a 90-day pre-release window on frontier AI models. The new order trims that window to 30 days, makes participation voluntary, and explicitly forecloses the regulatory architecture safety hawks have been pushing for since 2023.

The mechanism is narrow by design. Developers of “covered frontier models” may hand systems to Treasury, NSA, and CISA for classified benchmarking before release to trusted partners. Treasury has 30 days to stand up an AI cybersecurity clearinghouse to coordinate vulnerability scanning and patch distribution with industry. Section 3 governs designation; the operative legal posture sits in the disclaimer: “Nothing in this section shall be construed to authorize the creation of a mandatory governmental licensing, preclearance, or permitting requirement.”

That sentence is the entire policy. Per NPR, any binding regime has to come from Congress, and the White House isn’t pretending otherwise.

The proximate trigger, per NPR, was Anthropic’s April disclosure that its Mythos Preview model could autonomously find and exploit software vulnerabilities. CFR frames Mythos as the backdrop for the classified NSA-run designation process. Capability disclosures from one lab now set the tempo for federal action.

Trump postponed the May draft citing “certain aspects” he disliked and worries about US AI jobs, per Roll Call. The 12-day turnaround produced a document that does less.

Sen. Josh Hawley (R-Mo.) endorsed the order and immediately positioned past it: “I think we ought to enact my legislation that I have with Senator [Richard] Blumenthal that would make that sort of reporting and monitoring mandatory.” The endorsement reads as a marker. Voluntary today, statutory tomorrow, if a frontier model embarrasses someone first.

Sources

Sources