OpenAI's robotics lead walks out over defense ethics

OpenAI just learned that winning a Pentagon deal comes with internal costs.

Caitlin Kalinowski, the company’s robotics hardware lead, has publicly resigned in protest over OpenAI’s controversial defense contract. In a post on X, she stated that the agreement moved forward without clearly defined guardrails around surveillance and lethal autonomy.

Her departure marks the first senior-level exit explicitly tied to the deal. It follows a chaotic week that began when Anthropic walked away from military negotiations, got slapped with a supply-chain risk designation, and saw OpenAI rush in to take the contract instead.

Walking away on principle

Kalinowski joined OpenAI in November from Meta, where she previously led hardware for augmented reality glasses. Her arrival signaled OpenAI’s serious effort to rebuild its robotics division, which had been shut down in 2020.

In her announcement, Kalinowski made it clear that her decision was about principle rather than personal animosity. She noted that while AI plays an important role in national security, surveillance of Americans without judicial oversight and lethal autonomy without human authorization are lines that deserved far more deliberation.

In a follow-up post, she added that the announcement was rushed without defining guardrails, calling it primarily a governance failure. An OpenAI spokesperson confirmed her departure, maintaining that the agreement establishes a workable path for defense AI while explicitly banning domestic surveillance and autonomous weapons.

Backlash on two fronts

The discontent inside OpenAI is already mirroring user reactions outside the building. While research VP Max Schwarzer also left for Anthropic last week, consumers have staged their own walkout: ChatGPT uninstalls surged 295% following the defense announcement.

Anthropic quickly capitalized on the sentiment. Claude shot to the top of the U.S. App Store free charts, relegating ChatGPT to the second spot, while major cloud partners like Microsoft, Google, and Amazon pledged continued support for Claude across non-defense workloads.

Why it matters

For users, AI provider selection is rapidly shifting from benchmark scores to corporate values. If you are an everyday developer or creator, swapping between Claude and ChatGPT has minimal technical friction, but many users clearly prefer not to subsidize defense-contract models when comparable alternatives exist.

Within the industry, the contract scramble creates a stark divide between Anthropic’s hardline stance and OpenAI’s rapid accommodation of military demands. While OpenAI claimed it had multi-layered safeguards, watching its robotics director walk out suggests those internal guardrails exist mostly on paper right now.

We think OpenAI made an unforced error by treating national defense contracting like a standard tech landgrab. Rushing an announcement to capitalize on a rival's standoff might satisfy business development goals, but it actively alienates top-tier talent. When the person hired to put AI into physical machines quits over concerns about autonomous weapons, it indicates that internal governance is struggling to keep pace with executive dealmaking.