OpenAI robotics chief resigns after Pentagon deal

Caitlin Kalinowski resigned as OpenAI’s head of robotics on Saturday, saying she could not support the company’s new deal with the Pentagon. In a post on X, Kalinowski wrote that national-security work for AI can be legitimate, but drew two lines she said OpenAI crossed too quickly: “surveillance of Americans without judicial oversight” and “lethal autonomy without human authorization”, according to Business Insider.

OpenAI confirmed her resignation and defended the agreement with the US Defense Department. A company spokesperson told Business Insider that the deal creates “a workable path” for responsible use while setting “red lines: no domestic surveillance and no autonomous weapons.” The dispute is less about whether such red lines can be written down than about who enforces them once the customer is a national security bureaucracy and the supplier is a private firm competing for contracts.

The timing matters. OpenAI signed its Pentagon deal last week after rival Anthropic refused a similar arrangement, citing concerns about mass surveillance and autonomous weapons. Within days, Anthropic was hit with political retaliation: Trump called the company “radical woke” and demanded agencies stop using its technology, and defense secretary Pete Hegseth designated Anthropic a supply-chain risk and barred Defense Department contracts from working with it, Business Insider reported.

That sequence creates a simple lesson for vendors: declining defense work can be punished as disloyalty, while accepting it can be framed as “responsible engagement” as long as the marketing language is careful. In that environment, internal dissent becomes a reputational variable to manage. A high-profile resignation can function as a pressure valve—signalling to employees and users that objections were raised—without changing the underlying commercial trajectory.

The market response described by Business Insider was immediate. Some users said they were ditching ChatGPT in protest, while Anthropic’s Claude rose to the No. 1 free app in the US Apple App Store, unseating ChatGPT. Business Insider reported that Claude’s US downloads jumped 240% month-on-month in February.

Kalinowski joined OpenAI from Meta in 2024 and led its robotics division, a role that sits close to the boundary between software and physical systems. As defense agencies increasingly buy “capability” as a subscription—models, updates, hosting and access controls—questions about who can switch a system off, and under what legal authority, become operational rather than philosophical.

Kalinowski’s resignation was posted on X. OpenAI’s Pentagon partnership remains in force.