President Donald Trump delayed signing an executive order on AI security evaluation on May 21, scrapping a planned White House ceremony after objecting to proposed language requiring AI companies to share advanced models with government agencies 14 to 90 days before public launch. Trump told the White House press pool he was concerned the order would hinder US competitiveness against China.
The anticipated order would have tasked the Office of the National Cyber Director and other federal agencies with developing a formal process to evaluate AI models for security risks before their commercial release. The pre-release sharing requirement represented the order's most contentious provision, with Trump calling the language a potential blocker to American AI leadership.
“We're leading China, we're leading everybody, and I don't want to do anything that's going to get in the way of that leading.”— Donald Trump, President
The timing follows the recent releases of Anthropic's Mythos and OpenAI's GPT-5.5 Cyber, both models capable of rapidly identifying and exploiting security vulnerabilities in software systems. Those capabilities prompted renewed debate inside the administration about balancing innovation velocity against national security oversight.
Key facts
- 01Trump delayed an executive order requiring AI companies to share advanced models with government 14 to 90 days before launch.
- 02The order would have tasked the Office of the National Cyber Director with developing pre-release security evaluations.
- 03Trump cited concerns about maintaining US competitiveness against China as reason for the delay.
- 04The order was drafted partly in response to Anthropic's Mythos and OpenAI's GPT-5.5 Cyber security capabilities.
Multiple reports indicated insufficient CEO availability for a May 21 signing ceremony contributed to the delay, though Trump's stated objections centered on the order's substance rather than scheduling. The president said he did not like certain aspects of the draft language as written.
“could have been a blocker”— Donald Trump, President
The proposed 14-to-90-day advance sharing window would have given government evaluators time to assess frontier models before they reached public deployment. Supporters argued the window was necessary to identify catastrophic security risks; critics inside the AI industry have consistently opposed mandatory pre-release government review as a regulatory bottleneck.
Trump framed the delay as a competitiveness decision rather than a security retreat. He emphasized that the US currently leads China and other nations in AI development and said he would not sign measures that jeopardize that position.
The Office of the National Cyber Director has been developing AI security evaluation frameworks for the past 18 months but has not yet published formal guidelines for pre-release model assessments. The executive order would have accelerated that timeline by mandating a government review process.
No revised draft or rescheduled signing date has been announced. The administration has not specified which aspects of the language require revision or whether the 14-to-90-day sharing requirement will remain in future versions.
Industry reaction to the delay has been muted publicly, though AI companies have privately lobbied against mandatory pre-release disclosures since similar provisions appeared in earlier legislative proposals. The lack of a formal review process leaves current practice unchanged, with companies voluntarily sharing some model details through existing government partnerships.
The delay creates a regulatory gap at a moment when offensive AI capabilities are advancing rapidly. GPT-5.5 Cyber and Mythos both demonstrated autonomous exploit-finding at speeds previously requiring human security researchers, raising questions about whether existing voluntary frameworks can keep pace with capability growth.
Trump's competitiveness framing signals that future AI regulation from this administration will prioritize speed-to-market over precautionary review. That posture aligns with the White House's broader stance on technology policy but leaves unanswered how the government will evaluate security risks from models capable of autonomous cyberattacks before those models reach adversaries.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
