By Hadas Gold, CNN
(CNN) — President Donald Trump on Tuesday signed an executive order seeking early government access for the most advanced AI models to weigh cybersecurity risks and protect critical infrastructure.
The executive order comes as models like Anthropic’s Mythos have spooked government and Wall Street over security concerns.
The order requests AI companies voluntarily share new models deemed to have advanced cyber capabilities with the government up to 30 days before providing access to other partners.
Trump was expected to introduce the executive order nearly two weeks ago, but it was postponed just hours before the planned signing ceremony.
An earlier draft of the executive order from May viewed by CNN called for a longer 90-day review period. Multiple sources in the AI industry told CNN many AI companies preferred a shorter review period considering how rapidly AI models are developing.
A meeting on Monday with top Trump advisers and cabinet officials helped push the new version through, according to a source familiar with the situation .
A White House spokesperson said the executive order reflects Trump’s “common-sense approach of collaborating with industry to balance innovation and security, cementing America’s continued global dominance in AI and cybersecurity.”
The order also calls on national security agencies to improve their cybersecurity defense with a “cybersecurity clearinghouse.”
Major AI companies engaged directly with the White House on developing the executive order. That includes Anthropic, which the Pentagon blacklisted and deemed a “supply chain risk” after a disagreement over guardrails in its models on classified military systems.
“This executive order is an important step toward advancing innovation while protecting the security of the American public. We welcome this effort by the Administration,” Microsoft President Brad Smith said in a statement.
The Trump administration had taken a more hands-off approach to AI regulation until Anthropic unveiled its Mythos model, which it says can exploit cybersecurity vulnerabilities at an unprecedented pace.
Advanced AI models can supercharge cyberattacks, experts have said, and an early review could help the government protect against threats before they’re unleashed in the world.
Anthropic and OpenAI have given select companies and governments access to their most advanced models to help them prepare their defenses. Last week Anthropic also announced that its Mythos model will be released “to all our customers in the coming weeks.”
The Department of Commerce’s National Institute of Standards and Technology announced last month that major tech companies will share unreleased versions of their AI models with the government for national security and public safety-related evaluation. But that announcement is no longer available on the Commerce Department’s website.
The original executive order was initially supposed to be signed at a hastily arranged ceremony at the White House on May 20. Executives from the top AI and tech firms had been invited, and some were en route to the White House when the order was postponed.
Trump told reporters that same day he had delayed the EO signing because he “didn’t