The Trump administration is pressing Meta to submit its artificial intelligence models for voluntary review, which would allow the government to evaluate the A.I.’s abilities and vulnerabilities, four people familiar with the confidential request said.
The request, which was made in emails with Meta, is the latest example of the administration’s efforts to step up oversight of the A.I. industry after promoting a hands-off approach to the powerful technology. Less than two weeks ago, the government ordered Anthropic to remove access to its newest model, citing national security concerns.
Meta is the only major U.S. developer of A.I. technology that has not reached an agreement to voluntarily share its models with the federal government for review, said the people familiar with the request, who spoke on the condition of anonymity because they were not authorized to discuss the matter publicly. OpenAI, Anthropic, Google, xAI and Microsoft have all agreed to submit their models to the government’s A.I. safety group, known as the Center for A.I. Standards and Innovation.
A Meta spokesperson did not immediately provide a comment on the government’s request.
Ben Kass, a Commerce Department spokesman, said the Center for A.I. Standards and Innovation, which is housed in the department, regularly engaged with companies about voluntary agreements.
“This story is not unusual,” he said. “It is the very work CAISI is supposed to be doing.”
The latest batch of A.I. models have increased concerns about cybersecurity, though some industry insiders say the fears are overblown, since the new technology can be used to defend computer networks as easily as it can be used to attack them.
The companies have been submitting their A.I. models for review for several months as a good-will gesture, so intelligence and defense officials could ensure that the latest A.I. products did not pose a risk to national security, two government officials said.
On June 2, President Trump signed an executive order that gave the government responsibility for A.I. reviews. The order asked tech companies to give the U.S. government up to 30 days to evaluate A.I. models before their release to the public. It gave the government until the end of July to develop a process for the reviews.
But it is unclear who will lead the efforts and what type of standards the models would be held to, one person involved in the process said.
The Center for A.I. Standards and Innovation has stepped up its efforts to play a role in the model review process. Overseen by Commerce Secretary Howard Lutnick, the agency was created by the Biden administration to vet A.I. models and has a technical staff to lead those evaluations.
Meta released its latest A.I. model, Muse Spark, in April, and it nearly matched the performance of models from rivals like Google, OpenAI and Anthropic.
Meta’s policy team has been negotiating with the Commerce Department about how to proceed, the people familiar with the confidential request said. It’s unclear whether they will be able to reach an agreement.
Even companies that have given the administration previews of their A.I. models have run into problems. Anthropic, which gave its newest and most powerful model, Fable 5, to the administration for review, was surprised last week when the White House gave the company less than 90 minutes to close access to its new A.I. because of national security concerns.
The problem, two people with knowledge of discussions said, stemmed from a paper written by researchers at Amazon that showed a vulnerability in the model that could be exploited for cyberattacks.
In the days since, talks between Anthropic and the administration have been productive, two people with knowledge of the talks said. Mr. Trump made comments over the weekend to Axios that he no longer saw Anthropic as a security concern.
It was unclear who would sign off on Anthropic’s restoring access to its model and whether other companies would be held to the same standard, the two people said.
Ana Swanson contributed reporting from Washington.
The post U.S. Presses Meta to Agree to A.I. Reviews as Security Concerns Rise appeared first on New York Times.
