Open AI CEO Sam Altman speaks during a talk session with SoftBank Group CEO Masayoshi Son at an event titled “Transforming Business with AI” held in Tokyo on February 3, 2025.
Tomohiro Osumi | Getty Images
OpenAI CEO Sam Altman told employees during an all-hands meeting Tuesday that the company “cannot make operational decisions” about how its artificial intelligence technology is used by the Department of Defense.
“So maybe they think the attack on Iran was good and the invasion of Venezuela was bad,” Altman said Tuesday, according to a partial transcript of the conversation seen by CNBC. “There’s no need to consider that.”
The meeting came four days after OpenAI announced the Pentagon deal, which arrived hours before the US and Israel launched attacks against Iran.
Altman told employees that the Pentagon respects OpenAI’s technical expertise, wants input on where its model fits, and will allow the company to build the security stack it sees fit, said the people, who requested anonymity because the meeting was private.
But Altman said the agency has also made it clear that operational decisions rest with Secretary Pete Hegseth. Altman has been loudly criticized by some OpenAI employees since announcing the partnership with the Department of Defense shortly after rival Anthropic was blacklisted and labeled a “supply chain risk to national security.” President Donald Trump also directed all federal agencies in the United States to “immediately cease” all use of Anthropic’s technology.
Anthropic’s AI was also reportedly used in last weekend’s attack on Iran and in January’s capture of ousted Venezuelan leader Nicolas Maduro and his wife, Cilia Flores.
Altman has defended OpenAI’s deal in various social media posts, but acknowledged that it “looks opportunistic and sloppy” and that the company “should not have rushed to announce the deal on Friday.” In a post on X the same day, he said the Pentagon had “demonstrated a deep respect for security and a desire to work together to achieve the best possible outcome.”
Anthropic was the first lab to deploy the model across the Pentagon’s classified network and was trying to negotiate the terms of an ongoing contract before talks broke down. The company wanted assurances that its model would not be used for fully autonomous weapons or mass surveillance of American citizens, while the Pentagon wanted Anthropic to agree to allow the military to deploy the model for all lawful uses.
Last year, OpenAI won a $200 million contract from the Department of Defense, allowing the agency to begin using the startup’s model in unclassified use cases. The new arrangement will allow the company to deploy the model across the department’s sensitive network.
Elon Musk’s xAI also agreed to deploy the model across classified use cases.
“We believe that we can come up with the best model that the government will be willing to work with, even if our safeguards cause problems for them,” Altman said Tuesday. “But there’s going to be at least one more adversary, and I think that’s xAI. xAI will effectively say, ‘We’ll do whatever you want.'”
Mr. Altman and Mr. Musk, two of OpenAI’s co-founders, are engaged in a bitter legal battle, scheduled to go to trial next month.
XAI did not respond to requests for comment.
—CNBC’s Kate Rooney contributed to this report.
WATCH: OpenAI could make big money from deal with Pentagon, says Big Tech’s Alex Kantrowitz
