Trump's AI Order Makes Pre-Release Model Review Voluntary
· TexaHub · Regulation & Ethics, LLMs & Foundation Models, Geopolitics & Tech
Executive Order 14409 invites frontier AI developers to give the government up to 30 days' early access to cyber-capable models, but explicitly rules out mandatory preclearance.
What the order does
President Trump signed Executive Order 14409, "Promoting Advanced Artificial Intelligence Innovation and Security", on 2 June [1]. Its core provision gives the Treasury, the National Security Agency and the Cybersecurity and Infrastructure Security Agency, with other agencies, 60 days to develop a classified benchmarking process for the "advanced cyber capabilities of AI models" and to set the threshold at which a model becomes a "covered frontier model", a determination assigned to the NSA director [1].
Developers of such models would be able, under a voluntary framework, to give the federal government access for "a period of up to 30 days" before releasing them to other trusted partners, and to work with the government on which partners receive early access [1]. The order states that nothing in it authorises "a mandatory governmental licensing, preclearance, or permitting requirement" for new AI models [1].
Other provisions set 30-day deadlines for agencies to prioritise cyber defence of federal systems and for the Treasury to form an AI cybersecurity clearinghouse with industry and critical infrastructure operators to coordinate vulnerability scanning and patching [1].
A narrower order than first drafted
The signed text is lighter than earlier versions. The Guardian reported that an earlier draft reportedly requested models 90 days before release, and that Trump postponed signing in late May, saying on 21 May: "I didn't like certain aspects of it, I postponed it" [2]. The Next Web described the earlier draft as a 90-day mandatory review, a characterisation the Guardian presented only as a rumoured feature [2] [3].
The cyber focus reflects concern over models such as Anthropic's Claude Mythos, which the Guardian said has raised alarm for its ability to find exploitable software vulnerabilities [2]. Anthropic said on 28 May that Mythos-class models "require stronger cyber safeguards before they can be generally released" and that it expects to offer them to all customers "in the coming weeks" [4]. The Guardian also noted that Microsoft, Google DeepMind and xAI agreed last month to early government review of new models, and that the Commerce Department's Center for AI Standards and Innovation already has similar arrangements with OpenAI and Anthropic [2].
In practice, the order formalises arrangements that leading labs already accept, while leaving participation optional. Its effect will depend on where the NSA sets the classified threshold and whether developers outside the existing agreements choose to take part, neither of which is yet known.
Sources
[1] Promoting Advanced Artificial Intelligence Innovation and Security — The White House, 2 Jun 2026
[2] Trump signs executive order seeking early access to new AI releases — The Guardian, 2 Jun 2026
[3] Trump quietly signs a downsized AI executive order asking companies to voluntarily submit models for review 30 days before release — The Next Web, 2 Jun 2026
[4] Introducing Claude Opus 4.8 — Anthropic, 28 May 2026