OpenAI to amend Pentagon AI deal to bar domestic surveillance, freezes intelligence agency use

OpenAI is set to amend its agreement with the US Department of War (DoW) to explicitly prohibit the use of its artificial intelligence systems for domestic surveillance of Americans and to bar deployment to military intelligence agencies, including the National Security Agency, Chief Executive Sam Altman said in a post on X.

The move follows public scrutiny over OpenAI’s decision last week to deploy its models on the DoW’s classified networks and escalating tensions between the Pentagon and rival AI firm Anthropic over similar surveillance concerns.

Altman said OpenAI would add language to its contract stating that, “Consistent with applicable laws, including the Fourth Amendment to the United States Constitution, National Security Act of 1947, FISA Act of 1978, the AI system shall not be intentionally used for domestic surveillance of US persons and nationals.”

The amendment further clarifies that the limitation “prohibit[s] deliberate tracking, surveillance, or monitoring of US persons or nationals, including through the procurement or use of commercially acquired personal or identifiable information,” he wrote.

“It’s critical to protect the civil liberties of Americans,” Altman said, adding that the company wanted to make the restriction “especially clear, including around commercially acquired information.”

Altman also said the Department affirmed that OpenAI’s services “will not be used by Department of War intelligence agencies (for example, the NSA). Any services to those agencies would require a follow-on modification to our contract.”

The company’s decision to formalise the constitutional language comes days after Anthropic said it was facing a potential “supply chain risk” designation from the DoW after negotiations broke down over two requested exceptions: mass domestic surveillance and fully autonomous weapons.

In his post, Altman said he had reiterated in weekend conversations that Anthropic “should not be designated as a SCR, and that we hope the DoW offers them the same terms we’ve agreed to.”

OpenAI researcher Noam Brown, who led the company’s reasoning model breakthrough last year, said in a separate post that deployment to the NSA and other DoW intelligence agencies would be withheld “so that there’s time to address potential surveillance loopholes through the democratic process.”

“Over the weekend it became clear that the original language in the OpenAI / DoW agreement left legitimate questions unanswered, especially around some novel ways that AI could potentially enable legal surveillance,” Brown wrote. “The world should not have to rely on trust in AI labs or intelligence agencies for their safety and security.”

Brown added that he planned to become more personally involved in policy discussions at OpenAI, saying that as AI becomes more powerful, “it’s more important than ever that ultimate authority be vested in the public.”

Altman acknowledged that OpenAI “shouldn’t have rushed” its initial announcement, calling the issues “super complex” and saying clearer communication was needed.

He also said that if he received what he believed was an unconstitutional order, “of course I would rather go to jail than follow it.”

The developments underscore growing friction between frontier AI companies and US defence officials as governments seek access to advanced AI systems for national security purposes, while companies attempt to set guardrails around surveillance and the use of force.

OpenAI said it would host an all-hands meeting to address employee concerns and continue refining safeguards through what Altman described as an “iterative deployment” approach.