Trump signs executive order to review AI models before release

voluntary pre-release – President Donald Trump signed an executive order Tuesday creating a voluntary framework for AI companies to share frontier models with the federal government before release, aiming to assess advanced cyber capabilities and improve defenses for critical infrast

By the time President Donald Trump signed the executive order on Tuesday, it carried a very specific promise: AI frontier models would be checked for cyber risk before they hit the public.

The order creates a “voluntary framework” for AI companies to share their models with the federal government before they’re released. framed as a way “to promote secure innovation and strengthen the cybersecurity of critical infrastructure.” It leans on a familiar argument about speed—stating that the U.S. AI industry has advanced in part because the country “refuse[s] to stifle this innovation with overly burdensome regulation”—but it also spells out the concern that new AI capabilities can introduce new security risks.

To address that. the order directs multiple federal agencies to produce a framework that would “assess the advanced cyber capabilities of AI models” before they go public. Companies would decide whether to share their models pre-release, but the government would offer certain confidentiality protections if they do.

The order also requires the federal government to prepare cyber defenses for AI, with special emphasis on critical infrastructure.

This is not the first draft of the idea. Trump had postponed at the last minute signing a previously planned executive order that he said he worried could “get in the way” of competing with China. That earlier version would have let AI companies voluntarily share their models 14 to 90 days before release. In the current version. the timing tightens: companies would share models up to 30 days before public release. according to details previously reported.

In parallel to Washington’s evolving posture. major parts of the AI industry have already been testing the idea of pre-release review. Google. Microsoft. and xAI agreed last month to allow pre-release review by the Commerce Department’s Center for AI Standards and Innovation (CAISI). Separately. OpenAI and Anthropic had already agreed to share their models with CAISI back in 2024 under President Joe Biden. as part of Biden’s push for AI safety guardrails.

Still, the order tries to draw a line around what it is—and what it is not. It explicitly says it shouldn’t be taken as a form of mandatory licensing or preclearance. Even so. it reflects a willingness by the Trump administration to exercise oversight over AI companies in this new cybersecurity-focused way.

The shift arrives after a period when the Trump administration had downplayed safety concerns and taken a hands-off approach under former White House AI czar David Sacks. The language signed Tuesday suggests that stance has moved.

One possible spark sits inside Anthropic’s recent rollout history. The order points to a moment Anthropic has described publicly: the company’s limited April rollout of its Mythos model. Anthropic said Mythos flagged “thousands of high-severity vulnerabilities, including some in every major operating system and web browser.”.

For Anthropic, Mythos has also been tied to business and political friction. The same model appeared to create an opening to thaw tensions between Anthropic and the administration. following Anthropic’s legal battle with the Pentagon over its use of AI for autonomous lethal weapons and mass surveillance.

Not everyone who opposes broad state-level AI restrictions is cheering for tighter rules in general. but at least some see this order as a workable middle ground. Americans for Responsible Innovation President Brad Carson praised the decision. saying. “The White House is officially Mythos-pilled.” In a statement. Carson said the order shows the Trump administration is taking AI vulnerabilities seriously.

Alliance for Secure AI CEO Brendan Steinhauser offered similar support. He said his group is “pleased to see that the Trump administration is taking the risks of these models seriously,” and both Steinhauser and Carson urged Congress to codify mandatory protections.

There’s a clear tension baked into the order itself: it emphasizes voluntary participation and avoids licensing or preclearance. while still directing agencies to assess AI models’ advanced cyber capabilities and requiring government cyber defenses—especially for critical infrastructure. The result is a form of oversight that companies can opt into. but which signals that cybersecurity review is becoming part of the AI release conversation whether the industry feels ready for it or not.

Trump executive order AI models voluntary framework CAISI cybersecurity critical infrastructure frontier models Anthropic Mythos OpenAI Microsoft Google xAI