Caitlin Kalinowski, OpenAI’s hardware lead overseeing the company’s robotics initiative, resigned on March 7, 2026, directly citing the company’s newly announced partnership with the U.S. Department of Defense. Kalinowski, who joined OpenAI in November 2024, stated that the deal was announced before adequate guardrails were established, calling the announcement “a governance concern first and foremost.” Her departure marks a significant internal fracture at one of the most closely watched companies in the AI industry.
What Kalinowski Said and Why She Left
Kalinowski announced her resignation via social media, framing her decision as a matter of principle rather than a conflict with leadership. “Surveillance of Americans without judicial oversight and lethal autonomy without human authorization are lines that deserved more deliberation than they got,” she wrote. She emphasized having “deep respect” for CEO Sam Altman and her former colleagues but said she could not remain after the Pentagon deal was made public without the limits she considered essential being defined in advance.
Before joining OpenAI, Kalinowski spent years at Oculus working on virtual reality hardware and previously designed MacBooks at Apple. She then led the team that built Meta’s Orion augmented reality glasses. At OpenAI, her mandate was to build out the company’s robotics organization as it expanded into physical AI systems. Her resignation removes one of the company’s most experienced hardware leaders during a period when OpenAI has been publicly ramping up its ambitions in robotics and physical infrastructure.
What the Pentagon Deal Entails
The agreement allows OpenAI to make its AI models available within secure, classified computing environments operated by the Department of Defense. OpenAI responded to the controversy by stating the partnership “creates a workable path for responsible national security uses” and that the company maintains firm “red lines: no domestic surveillance and no autonomous weapons.” The company did not clarify what enforcement mechanisms would ensure those limits hold in practice, or what recourse would exist if the terms were not followed.
The context surrounding the deal added to its public scrutiny. Anthropic had previously been in discussions with the Pentagon to deploy its AI models but rejected similar terms after the Defense Department declined to include explicit restrictions on domestic surveillance and fully autonomous weapons systems. OpenAI subsequently reached an agreement on conditions that Anthropic had turned down. CEO Sam Altman acknowledged in remarks that the deal’s rollout appeared “opportunistic,” a concession that suggested awareness inside the company about how the sequence of events would be perceived externally.
The Broader Landscape for AI and Defense Contracts
The OpenAI Pentagon deal is part of an accelerating trend of commercial AI companies entering into formal arrangements with the U.S. national security apparatus. Defense contractors and technology firms have long pursued government contracts, but the pace has intensified as large language models move from research environments into operational use. The federal government has made clear its intent to integrate advanced AI across defense, intelligence, and security applications, and companies with leading models are under pressure to participate or risk competitors locking up access.
This is not the first time a major technology startup has navigated the intersection of defense partnerships and workforce ethics. Palmer Luckey, who founded Anduril after his exit from Oculus, built his second company entirely around defense technology and has secured more than $1.5 billion in funding to scale AI-driven defense manufacturing. Anduril represents one model, a company that chose from day one to operate at the intersection of AI and military contracting. OpenAI, which began as a research nonprofit with an explicit safety mission, represents a different and more contested trajectory.
What This Means for Founders and the Startup Ecosystem
For founders building on OpenAI’s API or competing in adjacent AI markets, the deal has tangible implications. OpenAI is now an active government defense contractor, and that status changes the risk calculus for startups that depend on its platform. Questions around data security, model access, and the future direction of OpenAI’s research priorities take on new weight when one of the company’s major customers is the U.S. military.
Kalinowski’s resignation also illustrates a dynamic that more founders are likely to encounter as AI capabilities attract government interest: the moment when a company’s mission and its commercial strategy diverge far enough to prompt key people to leave. Her departure was public, specific, and grounded in concrete policy objections, not vague discomfort. That specificity is notable. It signals that governance failures in AI, specifically the gap between when a deal is announced and when safeguards are defined, are increasingly becoming reasons that experienced operators choose to walk away. As top investors have noted, the ability to understand regulatory and ethical landscapes is becoming a core competency for founders building in AI.
OpenAI has not indicated whether it will revise the terms of the Pentagon agreement in response to internal or external criticism. As of March 9, 2026, the deal remains in place and the company continues to expand its government-facing operations. The situation continues to develop.
