Trump signs AI security review order with shorter vetting window than draft

President Donald Trump signed an executive order Monday requiring AI technology companies to submit their most advanced models to the government for security vetting before release.

The White House said the order is intended to consolidate U.S. dominance in AI while protecting critical infrastructure and national security systems from cyber threats.

At its core, the order establishes a government framework for pre-release security reviews of high-performance AI models developed by private companies.

The order largely mirrors a draft Trump had been expected to sign on May 21 before pulling back at the last minute, but with at least one significant change: the maximum security review window has been cut from 90 days to 30 days.

Under the order, the Treasury Department, the Defense Department and the Department of Homeland Security will establish a classified benchmarking process, while developers will set up voluntary cooperation frameworks with the government.

Developers will determine whether their models meet the threshold for review, and the government may access a model for up to 30 days before its release to check for security vulnerabilities.

A voluntary AI cybersecurity partnership to be formed under the order will scan for and identify software vulnerabilities at scale and coordinate the distribution of security patches.

The Trump administration said the framework differs from the top-down regulatory approach of the Joe Biden administration and does not introduce AI licensing or pre-approval requirements.

According to The New York Times, Trump had been set to sign the original draft on May 21 but canceled at the last moment after concerns were raised that a review period of up to 90 days was too long.

Tech figures including venture capitalist Marc Andreessen argued that while security checks were necessary, U.S. companies could not afford to fall behind China in the development race. David Sacks, co-chair of the President's Council of Advisors on Science and Technology, ultimately persuaded Trump to withdraw the original draft.

White House Chief of Staff Susie Wiles then led a renewed round of negotiations. At a closed-door meeting at the White House on Sunday, Trump, Sacks, Treasury Secretary Scott Bessent and Defense Secretary Pete Hegseth agreed to cap the review period at 30 days.

The Trump administration had resisted AI regulation until earlier this year, but shifted course after Anthropic developed a new AI model called Mythos in April. The model heightened concerns among national security officials and the financial sector over hacking and cyberattacks, prompting the push to establish a security review process.

The tech industry reacted with cautious relief after the order was announced Monday, while closely watching for its broader implications.

Google, Microsoft and OpenAI all welcomed the executive order.

OpenAI CEO Sam Altman is scheduled to visit the White House on Wednesday — a visit planned independently of the signing — to discuss follow-up measures with senior administration officials.

Sen. Mark Warner of Virginia, the ranking Democrat on the Senate Intelligence Committee, welcomed the move but offered a pointed criticism. He said the administration had "belatedly discovered the need to redo what they hastily dismantled in the first year of their term," according to the Associated Press.

Some voices are calling for legally binding review requirements rather than a voluntary framework.

Brendan Steinhauser, executive director of the advocacy group Alliance for Safe AI Action, told The New York Times that companies would ultimately comply even under a voluntary structure given the president's resolve, but added: "We want mandatory, legally enforceable review — not voluntary." He said the group would proceed as planned with a rally in Washington on Wednesday calling for stronger regulation.