President Donald Trump on June 2 signed an executive order titled Promoting Advanced Artificial Intelligence Innovation and Security, asking U.S. AI developers to voluntarily share new frontier models with the federal government for up to 30 days before public release and creating a government-wide push to assess the cyber capabilities of those models.
Voluntary 30-day government preview
The order asks technology companies to collaborate with the administration to "select trusted partners" that will gain early access to upcoming models in order to "promote secure innovation and strengthen the cybersecurity of critical infrastructure." Crucially, the order explicitly prohibits any mandatory licensing, preclearance or permitting requirement for releasing new AI models, framing the program as voluntary.
NSA-led benchmarking of covered frontier models
The order directs the development of a classified benchmarking process to assess the advanced cyber capabilities of AI models and determine which ones constitute a "covered frontier model," with the Director of the National Security Agency making the designation. Treasury, the Office of the National Cyber Director, DoD, NSA, DHS and CISA are also tasked with standing up an "AI cybersecurity clearinghouse" to coordinate vulnerability discovery and remediation with industry and infrastructure operators including power companies and hospitals.
A reversal from the laissez-faire posture
The signing marks a striking shift for the Trump White House, which had previously taken a more hands-off approach to AI than the prior Biden administration and had recently backed away from another proposed safety order, an order that Scientific American notes had been partly inspired by the release of Anthropic's powerful Mythos preview model.
Industry watches a voluntary regime form
The order arrives as governments and frontier labs negotiate a new social contract on AI safety, alongside Anthropic's Project Glasswing expansion to 150 organizations, the Anthropic confidential IPO filing and accelerating enterprise AI rollouts on governed data platforms.
Reporting based on coverage from The White House, Scientific American, JURIST, NPR and CNBC.