OpenAI has struck a deal with the Department of Defense to deploy its models across the agency’s networks, company head Sam Altman revealed on Twitter. In his post, he said the two most important security principles of OpenAI are “a ban on domestic mass surveillance and humanitarian responsibility for the use of force, including autonomous weapons systems.” Altman claimed that the company had laid out those principles in its agreement with the agency, which he called by the government’s preferred name, the Department of War (DOW), and that he had agreed to respect them.
Shortly after President Donald Trump ordered all government agencies to stop using the cloud and any other anthropogenic services, the agency has closed the deal with OpenAI. If you’ll recall, US Defense Secretary Pete Hegseth previously threatened that if he continued to refuse to remove the guardrails on his AI, he would be labeled an anthropic “supply chain risk”, preventing the technology from being used for mass surveillance against Americans and in fully autonomous weapons.
It’s unclear why the government agreed to work closely with OpenAI if its models have similar guardrails, but Altman said she is asking the government to offer the same terms to all AI companies it works with. Jeremy Levin, the senior official under secretary of state for foreign assistance, humanitarian affairs and religious freedom, said on Twitter that the DoW “references some existing legal authorities and incorporates some mutually agreed protection mechanisms” in its contracts. Both OpenAI and xAI, which had previously signed a deal to deploy Grok in the DoW’s classified systems, agreed to those terms. He said this was the same “compromise that was offered to Anthropic, and was rejected.”
Anthropic, which began working with the US government in 2024, refused to bow to Hegseth. In his latest statement, published just hours before Altman announced the OpenAI deal, he reiterated his stance. “No amount of threats or punishment from the War Department will change our position on mass domestic surveillance or fully autonomous weapons,” Anthropic wrote. “We will challenge any supply chain risk designation in court.”
Altman said in his post on X that OpenAI would create technical safeguards to ensure that the company’s models behave as they should, claiming that DOW wanted the same. It is sending engineers to work with the agency to ensure [its models’] Security,” and it will be deployed only on cloud networks the new York Times Note, OpenAI is not yet on the Amazon cloud, which is used by the government. But that may soon change, as the company has also announced a partnership with Amazon to run its models on Amazon Web Services (AWS) for enterprise customers.
<a href