Caitlin Kalinowski, OpenAI’s head of hardware and robotics engineering, recently announced that she has resigned from the company. She will lead OpenAI’s robotics team starting in November 2024. She posted on X and LinkedIn, "I resigned from OpenAI. I care deeply about the robotics team and the work we have accomplished together. This was not an easy decision. AI has an important role in national security. But monitoring Americans without judicial oversight and achieving lethal autonomy without human authorization are two red lines that should have received more careful discussion than they actually have."
She emphasized that her decision was "about principles, not individuals," and said she still has deep respect for Sam Altman and the team and is proud of the results we have achieved together.
Kalinowski's departure comes amid an escalating industry debate over the extent to which AI companies should support the U.S. military's use of their technology. Negotiations between the U.S. Department of Defense and Anthropic broke down recently over the company's insistence on strict restrictions on domestic surveillance and the use of autonomous weapons. Shortly after the negotiations fell through, OpenAI reached an agreement with the Pentagon to deploy its AI models on a classified government network. This move sparked criticism from some employees and observers, who believed that OpenAI was taking over after Anthropic rejected relevant conditions. In response, OpenAI CEO Sam Altman later admitted that the external presentation of the deal "looked opportunistic," and the company has since begun to further clarify the limitations of how its AI systems can be used by the military.
A spokesperson for OpenAI issued a statement when confirming Kalinowski's departure, saying that the company believes that the agreement reached with the Pentagon provides a feasible path for "responsible use of AI in the field of national security" and clearly draws the red line: not for domestic surveillance and not for autonomous weapons. The statement pointed out that the company recognizes that the public has strong differences on these issues and will continue to maintain dialogue with employees, governments, civil society and communities around the world.
Before joining OpenAI, Kalinowski had long been engaged in hardware product development at large technology companies. She served as a hardware executive at Meta for more than two and a half years, leading the augmented reality glasses project code-named "Project Nazare" and later officially named "Orion", a product that Meta calls "the most advanced AR glasses ever made." In September 2024, Meta demonstrated a prototype of this AR glasses. Before taking charge of the Orion project, she worked at Meta's Oculus for more than nine years, focusing on the development of virtual reality headsets. Prior to that, she worked at Apple for nearly six years, participating in MacBook product design, including Pro and Air models.
Kalinovsky's departure highlights the pull and disagreement over ethical boundaries within the field of AI regarding dual-use military and civilian use, especially surveillance and lethal autonomous weapons. On the one hand, there is strong demand for advanced AI capabilities from national security agencies, and on the other hand, there are continuing concerns among technicians and companies about privacy protection, human control, and the risks of automation in warfare. As negotiations between the U.S. Department of Defense and major AI companies advance, similar value conflicts and personnel flows may continue to reshape the technical routes and ethical boundaries of this emerging industry.