OpenAI’s head of robotics, Caitlin Kalinowski, has resigned over the company’s contract with the US Department of War, saying key safeguards around domestic surveillance and autonomous weapons were not adequately reviewed before the agreement was signed.

“Surveillance of Americans without judicial oversight and lethal autonomy without human authorization are lines that deserved more deliberation than they got,” Kalinowski wrote in a LinkedIn post. “This was about principle, not people,” she added, saying she retained “deep respect” for CEO Sam Altman and the team she had built with them.

Kalinowski is the most senior OpenAI employee to publicly break with the company over the Pentagon deal. Hundreds of OpenAI and Google employees had previously signed an open letter calling on their companies to maintain limits on AI use for mass surveillance and autonomous weapons.

What it signals

For enterprise buyers, the resignation points to a governance process concern that contract amendments alone cannot resolve, said Sanchit Vir Gogia, chief analyst at Greyhound Research.

“When a senior leader leaves while raising concerns about insufficient deliberation around surveillance safeguards or lethal autonomy, the discussion quickly shifts from personalities to process,” Gogia said. “Inside established governance environments, agreements of this nature normally pass through multiple layers of review before they are finalised. When those discussions appear to continue after a contract has already been announced, enterprises tend to interpret the situation as evidence that governance frameworks are still evolving.”

The episode also reinforces a broader shift in how enterprises evaluate AI vendors, Gogia noted. “Organizations evaluating multiple AI model providers rather than standardising on one, and incorporating additional governance documentation requirements into vendor contracts before approving large-scale deployments.”

The public backlash wherein Anthropic’s Claude surged to the top of Apple’s App Store as users uninstalled ChatGPT will not automatically translate into enterprise churn, said Abhishek Sengupta, vice president at Everest Group. “Public sentiment is often reactionary and not sticky,” he said. “Enterprise decision makers have another risk vector to assess when thinking of their AI stack. Expect national security guidelines to increasingly impact AI sourcing considerations, especially for geopolitically relevant economies.”

The deal and its critics

OpenAI signed the Pentagon agreement on February 27, hours after the Department of War designated rival Anthropic a supply-chain risk over its refusal to allow its models to be used for domestic mass surveillance or fully autonomous weapons. Altman announced the deal that evening, saying the Department of War had agreed to OpenAI’s red lines on surveillance and autonomous weapons. By Monday, he conceded it had been mishandled, saying on X it “looked opportunistic and sloppy.”

Under pressure, OpenAI revised the agreement to explicitly prohibit domestic surveillance of US persons and extended the prohibition to commercially acquired data, including geolocation and browsing history. Intelligence agencies such as the NSA were excluded from the contract’s scope, according to an updated post on OpenAI’s website.

Legal experts remained unconvinced. The Electronic Frontier Foundation, a nonprofit digital rights group, warned that “secret agreements and technical assurances have never been enough to rein in surveillance agencies.”

“When a technology vendor revises a government contract after public criticism, enterprise buyers rarely treat the amendment itself as the real reassurance,” Gogia said. “Risk teams start asking how those commitments are implemented, who enforces them, and what happens if interpretations change later.”

Anthropic back at the table

The broader dispute between the Pentagon and the AI industry remains unresolved. Anthropic CEO Dario Amodei has resumed negotiations with the Pentagon. The Information Technology Industry Council, whose members include Apple, Google, Nvidia, and OpenAI, has written to the Department of War Secretary Hegseth arguing that the supply-chain risk designation is typically reserved for foreign adversaries and could restrict government access to leading American technology.

Altman has publicly backed Anthropic’s reinstatement. “Anthropic should not be designated as a SCR, and that we hope the DoW offers them the same terms we’ve agreed to,” he wrote on X.