For creators, engineers, and artists building with frontier models, the latest executive order means the US government now secures a 30-day head start on accessing new AI systems before they hit the public market. This shift from a proposed 90-day window to a tighter timeline ensures that officials can vet powerful tools for security risks without stifling the rapid pace of development that keeps innovators moving forward.
The final deal
President Donald Trump signed the revised directive Monday night after senior advisers convinced him that indefinite delays were no longer viable. The White House team drafted the final language Tuesday morning, settling on a framework that grants federal agencies exclusive access to advanced models 30 days prior to public release.
While the review period was shortened, the core mechanics remained largely intact from the original proposal. The order does not create formal regulation but establishes a voluntary process to identify the most capable systems, allowing time to patch vulnerabilities before they become available to everyone.
Security concerns drive the push
The urgency stems from growing anxiety within the administration that systems like Anthropic‘s Claude Mythos and OpenAI’s GPT-5.5 could be weaponised for cyberattacks against critical infrastructure. Officials felt that waiting longer than necessary posed a tangible risk to national security.
The directive also tasks the Pentagon with reinforcing classified networks within 30 days and instructs the Justice Department to pursue criminal charges against individuals using AI models to hack computer systems.
A victory for the White House team
The move marks a significant win for White House Chief of Staff Susie Wiles, who partnered with Treasury Secretary Scott Bessent and National Cyber Director Sean Cairncross to resurrect the proposal. They overcame initial resistance from Trump’s former AI czar, David Sacks, a vocal critic of government intervention in the sector.
Following the signing, Bessent is now positioned to discuss a similar cross-border framework with China, talks that had been paused while domestic policy was settled. White House spokesperson Liz Huston described the order as a “common-sense approach of collaborating with industry to balance innovation and security, cementing America’s continued global dominance in AI and cybersecurity.”
Major players have already endorsed the direction. Anthropic posted on X that the order is an “important step in strengthening America’s leadership in AI” and expressed a willingness to collaborate on implementation. Previously, Trump had scrapped a version of this order on May 21 after companies warned that a 90-day window would be too burdensome for a fast-evolving industry. However, executives eventually argued that models were becoming too sophisticated to ignore, prompting the administration to act.
Behind the scenes, aides told several companies they expected a framework eventually but were unsure of the timeline. Trump ultimately felt there was sufficient industry buy-in for the 30-day window, giving his approval Monday night after a high-level meeting that included Wiles, Bessent, and a remote Sacks.
The order now requires federal agencies to establish a classified process for selecting which models warrant early access and to identify other “trusted partners” who may share in that access.
Key takeaways
“This Executive Order is an important step in strengthening America’s leadership in AI. We look forward to collaborating with the White House to support its implementation,” Anthropic wrote.
- The 30-day review window balances the need for security vetting with the practicalities of a rapidly advancing industry.
- Implementation includes directing the Pentagon to secure classified networks and empowering the Justice Department to prosecute AI-facilitated cyberattacks.
Stay ahead of AI. Get the most important stories delivered to your inbox — no spam, no noise.
