Trump plans to sign an executive order granting oversight of artificial intelligence models
President Trump plans to sign the executive order on Thursday giving people working on the order said the government has the authority to evaluate AI models before they are made public. It marks a shift for an administration that has promoted hands-free access to powerful technology.
The order will give the Office of the National Cyber Chief, which sits in the White House and oversees cybersecurity coordination across the government, and other agencies two months to develop a model evaluation process, the people said. The aim is for the government to identify any security vulnerabilities revealed by the AI models and fix problems in its systems to help protect banks, utilities and other sensitive industries from cyber attacks.
The White House has proposed that major AI companies voluntarily share their models 14 to 90 days before publication, the people said. The eventual process could also include creating a security vulnerability vault so that companies and cybersecurity researchers can report vulnerabilities they find using AI models, some of the people said.
Mr. Trump’s executive order would formally shift the White House away from its AI-based approach, which the president says could help advance the United States in the technology race against China. Now the administration plans to take a more hands-on stance. The decision to begin a formal oversight process stemmed from concerns that AI was becoming too powerful and could pose a security risk to the United States in the future, officials familiar with the discussions said.
Those concerns grew last month after startup Anthropic announced a new AI model, Mythos. Anthropic said the model could find software vulnerabilities and lead to a “cybersecurity showdown.” Government officials, banks and others worried that future AI models could find vulnerabilities that US enemies could exploit.
Those concerns pit U.S. officials, who favor some regulation of AI on national security grounds, against those who favor a hands-off approach to U.S. corporate interests, including AI companies. Until the release of Mythos, the Trump White House had largely ignored cybersecurity as a top policy priority.
“It’s pretty common not to fully understand the importance of cybersecurity until something massive happens,” said Cynthia Kaiser, former FBI top cybersecurity official and CEO of digital security firm Halcyon. “Mythos raised concerns in government, much like when a CEO discovers their own networks have suffered a massive breach.”
The White House did not immediately respond to requests for comment. Anthropic, OpenAI, Google and Microsoft also had no immediate comment.
Negotiations on the executive order began the first week of May, when White House officials met with AI officials. Officials had hoped Mr Trump would be able to sign the executive order before May 14, when he and Chinese leader Xi Jinping met for a summit in Beijing. The United States and China have found themselves in an AI arms race for everything from language models and chips to AI-powered weapons.
Within the Trump administration, some have argued that China’s control of domestic artificial intelligence has given Beijing an advantage by allowing it to control and access new models of artificial intelligence, two officials familiar with the discussions said. They suggested that the Trump administration could set up a regulatory body to review AI models before they are made public.
But officials ultimately decided the review board would create too much bureaucracy and slow down AI companies.
The US government previously relied on technology companies to voluntarily share information about their models. Not every AI company participated, and the organization responsible for the research, the Center for AI Standards and Innovation, published several reports on AI models released over the past year.
But after the release of Mythos, the White House faced pressure from banks and tech companies to do more. And public concern about AI has grown. In recent Quinnipiac University Survey 55 percent of American adults said they see AI as a force for harm rather than good, driven by concerns about job losses and mental health, as well as rising energy prices as data centers use more power.
On Tech Word News On May 6, Kevin Hassett, director of the White House’s National Economic Council, became one of the first administration officials to speak publicly about the AI evaluation process, saying the administration was considering evaluating the models “to, you know, release them into the wild after they’re proven to be safe, just like an FDA drug.”
His comments initially raised alarm in the tech industry, which embraced the government’s backtracking approach.
That night, White House Chief of Staff Susie Wiles posted her fourth social media post on X, saying the White House would “lead an America-first effort to empower great American innovators, not bureaucracy, to safely deploy powerful technologies while keeping America safe.”
The post gained administration time as he held discussions with major AI companies under the executive order. The model review process recalled some of the approaches once pushed by the Biden administration in 2024.
Mr Trump’s summit in China briefly interrupted the talks. When U.S. officials returned, they spoke with executives from Anthropic, OpenAI, Google and others to draft the wording of the executive order.
White House officials invited these executives to attend the signing of the executive order in the Oval Office on Thursday.
Dustin Volz contributed reporting.