OpenAI Secures Pentagon Contract After Anthropic Dispute, Raising AI Ethics Concerns
OpenAI claims it has accomplished what Anthropic couldn’t: securing a Pentagon contract that won’t cross professed red lines against dragnet domestic spying and the use of artificial intelligence to order lethal military strikes. Just don’t expect any proof.
The Deal and the Disagreement
Sam Altman, OpenAI’s CEO, announced the company’s contract with the Defense Department on February 27, 2026, via a post on X. Altman stated that two key safety principles – prohibitions on domestic mass surveillance and human responsibility for the use of force, including in autonomous weapon systems – are reflected in the agreement and align with the Pentagon’s laws and policies.
This deal followed the public breakdown of a similar contract between the U.S. Military and Anthropic, OpenAI’s main competitor. Anthropic’s negotiations failed because it could not guarantee prohibitions against autonomous weapons and domestic spying would be enshrined in the contract. This stance led to criticism from the Pentagon and President Donald Trump, who ordered a phase-out of Anthropic’s tools within six months.
OpenAI’s Approach and Lack of Transparency
OpenAI has attempted to justify its agreement by stating it focused on citing applicable laws rather than specific prohibitions in the contract, as Anthropic did. However, the company and the government have not released the contract itself, leaving the public unable to verify OpenAI’s claims.
The Department of Defense did not respond to requests for comment, and OpenAI representatives, including spokesperson Kate Waters, have only provided links to previous public statements from Altman when questioned about the contract’s specifics.
Concerns and Skepticism
Critics have raised concerns about the lack of transparency and the potential for the Pentagon to circumvent the stated limitations. Former military officials have expressed doubts about the contract’s language and its ability to prevent unwanted surveillance or the use of OpenAI’s technology in intelligence gathering. Brad Carson, a former under secretary of the Army, questioned whether the contract truly blocks Pentagon spy agencies like the NSA or the National Geospatial-Intelligence Agency from using OpenAI’s tools.
A former Pentagon official, speaking anonymously, highlighted the ambiguity of the term “intentional” surveillance, suggesting it could provide a loophole for the government to justify its actions. Alan Rozenshtein, a former Department of Justice National Security Division attorney, emphasized that without access to the contract, OpenAI’s claims are unsustainable and bizarre.
Questionable Statements and Past Actions
OpenAI’s public statements have also come under scrutiny. Katrina Mulligan, OpenAI’s national security chief, initially pledged to release a more comprehensive explanation of the contract’s terms but later stated she was not obligated to share the language. She also indicated a willingness to work with the NSA under the right conditions.
Mulligan incorrectly claimed the Pentagon lacks the authority to collect commercially available data, contradicting a declassified 2022 report from the Office of the Director of National Intelligence. Senator Ron Wyden has repeatedly highlighted the Pentagon’s surveillance of Americans’ data.
Trust and Accountability
The ultimate assessment of this agreement hinges on trust in the involved parties – OpenAI, Trump, and Defense Secretary Pete Hegseth – a trust that has been eroded by past actions and statements. Altman has faced accusations of dishonesty, and Hegseth’s record includes overseeing controversial military operations.
Without the contract’s text, it remains unclear whether OpenAI has genuinely secured meaningful safeguards or simply provided a public relations win for the Pentagon. The former Pentagon AI official stated that the arrangement is cause for alarm even with the most diligently crafted contract.