The head of OpenAI's robotics division has stepped down from the artificial intelligence firm after expressing significant concerns regarding a recent agreement with the US Department of War. Caitlin Kalinowski, who joined the ChatGPT creator in November 2024, stated her departure was fundamentally about principle, focusing on corporate governance and the urgent need to establish clear safety protocols around advanced technology.
Resignation Follows Controversial Government Agreement
OpenAI CEO Sam Altman publicly announced the partnership with the Pentagon on 27 February. This development occurred shortly after the Trump administration revealed intentions to terminate a contract with rival AI company Anthropic. The termination stemmed from Anthropic's refusal to comply with requests concerning the deployment of its Claude AI system for purposes like mass surveillance and autonomous weaponry.
Dario Amodei, Anthropic's chief executive, explicitly stated his company "cannot in good conscience accede to [the department's] request" for such applications. The new OpenAI deal subsequently triggered a backlash from dozens of current employees, who signed an open letter urging their company to adopt a stance similar to Anthropic's.
Principle Over Position: Kalinowski's Stance
Kalinowski announced her resignation via a LinkedIn post over the weekend, directly linking her decision to the manner in which the Department of War agreement was formulated and announced. She criticised the process as "rushed" and lacking the necessary definition of critical safety guardrails.
"This wasn't an easy call. AI has an important role in national security," Kalinowski wrote. "But surveillance of Americans without judicial oversight and lethal autonomy without human authorisation are lines that deserved more deliberation than they got." She emphasised that her future work will continue in the realm of physical artificial intelligence devices, with a steadfast focus on safety-first development.
OpenAI's Response and Revised Safeguards
In the wake of the resignation and internal criticism, Sam Altman acknowledged that the initial announcement of the Pentagon deal appeared "opportunistic and sloppy." He confirmed efforts to amend the agreement to incorporate additional protective measures.
OpenAI released a statement asserting that the revised contract language explicitly prohibits the use of its tools for domestic surveillance of US persons or for autonomous lethal weapons systems. The company claimed its agreement possesses more robust guardrails than any prior arrangement for classified AI deployments, including those of Anthropic.
"In our agreement, we protect our red lines through a more expansive, multi-layered approach," the statement detailed. "We retain full discretion over our safety stack, we deploy via cloud, cleared OpenAI personnel are in the loop, and we have strong contractual protections." This statement aims to address concerns about maintaining human oversight and ethical boundaries in government applications of AI technology.
