President Trump had planned on Thursday to sign an executive order that would give the government the power to evaluate artificial intelligence models before they were publicly released.
But hours before the event, the White House canceled the signing, even as executives from A.I. companies were midair on their way to the Oval Office.
Mr. Trump told reporters that he delayed the signing because he “didn’t like certain aspects of it,” without elaborating on what those aspects might be. He added, “I think it gets in the way of — you know, we’re leading China, we’re leading everybody, and I don’t want to do anything that’s going to get in the way of that lead.”
At the same time, A.I. companies were told Mr. Trump was not happy that many of their chief executives, who had been invited to the White House for Thursday’s signing, could not make it, people with knowledge of the conversations said. The White House had invited the leaders of OpenAI, Google, Anthropic, Meta and Microsoft just 24 hours earlier to the event. In lieu of the C.E.O.s who were tied up, some companies had flown other executives to the scheduled signing.
The cancellation capped weeks of back and forth over the executive order, which would have reversed the administration’s hands-off policy on A.I. The scramble showed how divided the White House has been over a fast-evolving and powerful technology that affects national security and the U.S. race against China.
The White House did not immediately respond to requests for comment.
The executive order would have granted the federal government oversight of new A.I. models before they were publicly released. In its current form, the order would give the Office of the National Cyber Director, which sits within the White House and oversees cybersecurity coordination in the government, and other agencies two months to develop a process for evaluating new A.I. models, the people said.
The goal was for the government to identify any security vulnerabilities revealed by A.I. models and to patch problems in its systems to help protect banks, utilities and other sensitive industries from cyberattacks.
The White House had also proposed that the major A.I. companies voluntarily share their models from 14 to 90 days before a public release, the people said. The final process could also include the creation of a vault for security vulnerabilities, for companies and cybersecurity researchers to report vulnerabilities they find using A.I. models, some of these people said.
The order had been scheduled to be signed Thursday afternoon. People who worked on the order said they expected it to be signed at a later date, though it was unclear if the wording and structure of the order would change.
The decision to begin a formal oversight process had stemmed from fears that A.I was becoming too powerful and could pose a security risk to the United States in the future, officials familiar with the discussions said.
Those fears increased last month after the start-up Anthropic announced a new A.I. model, Mythos. Anthropic said the model could find software vulnerabilities and lead to a cybersecurity “reckoning.” Government officials, banks and others worried that future A.I. models could find vulnerabilities that U.S. enemies would exploit.
Dustin Volz and Tony Romm contributed reporting from Washington.
