President Donald Trump signed an executive order on Tuesday creating a “voluntary framework” for AI companies to share their frontier models with the federal government before they are released “to foster secure innovation and strengthen the cybersecurity of critical infrastructure.”
The order says the US AI industry has succeeded in part “because we refuse to stifle this innovation with overly burdensome regulation”, but it also recognizes that new AI capabilities come with security risks. Accordingly, it directs several federal agencies to come up with a framework to “assess the advanced cyber capabilities of AI models” before releasing them to the public. Companies will have discretion whether or not to share their models with the government before release, but they may still have some privacy protections if they choose to do so. It also requires the federal government to prepare cybersecurity for AI, especially for critical infrastructure.
The order comes after Trump postponed at the last minute the signing of a previously planned executive order, which he worried could “get in the way” of competing with China. While the first version allowed AI companies to voluntarily share their models 14 to 90 days before release, according to the new York TimesThe current version asks companies to share their models up to 30 days before public release. Google, Microsoft and xAI agreed last month to allow pre-release review by the Commerce Department’s Center for AI Standards and Innovation (CAISI).
OpenAI and Anthropic had already agreed to share their models with CAISI in 2024 under President Joe Biden as part of Biden’s push for AI safety guardrails. But until recently, the Trump administration downplayed security concerns and took a pragmatic approach under former White House AI czar David Sachs. The order signed on Tuesday clearly states that this should not be construed as mandatory licensing or preclearance. Still, it shows some willingness on the part of the Trump administration to monitor AI companies.
“The White House is Officially Mythos-Pilled”
One factor in this change may be Anthropic’s limited rollout in April of its powerful Mythos model, which the company said had identified “thousands of high-severity vulnerabilities, including some in every major operating system and web browser.” Mythos also created an opportunity to ease tensions between Anthropic and the administration, following a legal battle with the Pentagon over the use of AI for autonomous lethal weapons and mass surveillance.
The newly signed order has so far received praise even from groups that have advocated for restrictions on state AI laws. “The White House is officially mythos-piled,” Brad Carson, president of Americans for Responsible Innovation, said in a statement. He said the order shows the Trump administration is taking AI vulnerabilities seriously. Brandon Steinhauser, CEO of the Alliance for Secure AI, 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.
<a href