Caitlin Kalinowski, who led OpenAI's robotics team, resigns over concerns about domestic surveillance and autonomous weapons after OpenAI's US DOD contract
Caitlin Kalinowski, who had been leading hardware and robotic engineering teams at OpenAI since November 2024, announced she has left the company.
FortuneSharon Goldman
Context & Ripple Effects
Kalinowski joined to lead robotics and consumer hardware after overseeing Meta's Orion program, making her departure consequential for an area where hardware choices can shape real-world AI deployment. Her exit follows earlier high-profile departures tied to OpenAI's safety posture, including Jan Leike's Superalignment resignation.
The immediate story is also part of a recurring internal-governance arc: coverage has previously described concerns that OpenAI was moving quickly on products and safety testing. The DOD contract brings that tension into the dual-use domain, where surveillance and weapons-related boundaries are especially salient.
First-order effects
OpenAI loses the leader responsible for hardware and robotics at a time when its defense relationship has made the intended uses of those capabilities a live internal issue.
Kalinowski's stated rationale puts domestic-surveillance and autonomous-weapons safeguards at the center of scrutiny around OpenAI's DOD work, rather than treating the departure as a routine executive change.
Second-order effects
OpenAI will face a harder retention and recruiting test for robotics and safety-adjacent talent if prospective hires view its defense commitments as insufficiently bounded; Kalinowski had been recruited specifically to build OpenAI's robotics and consumer-hardware effort.
Other AI labs pursuing government work may be pressed by employees and partners to specify operational limits on military and domestic-surveillance use, not merely broad safety principles.
Third-order effects
If similar departures persist, defense engagement could become a durable governance and labor constraint for frontier labs: commercial and state demand may advance together, while staff legitimacy depends on enforceable use boundaries.
The episode points toward dual-use AI governance becoming an organizational design problem—who can approve, audit, and refuse deployments—rather than only a model-safety question.
The trend: Frontier AI labs are becoming more state-compatible while facing sharper internal demands for credible limits on dual-use deployments.
I resigned from OpenAI. I care deeply about the Robotics team and the work we built together. This wasn't an easy call. AI has an important role in national security. But surveillance of Americans without judicial oversight and lethal autonomy without human authorization are …
@maylivesforever To be clear, my issue is that the announcement was rushed without the guardrails defined. It's a governance concern first and foremost. These are too important for deals or announcements to be rushed.
OpenAI's robotics head quit over surveillance and Skynet Thank you, Caitlin. I hope your courage inspires your colleagues. OpenAI is nothing without its people. [image]
notice how quickly narrative has changed from “evil gubment trying to create AI killbots” to “as it turns out we shouldn't give polyamorous “AI safety” commissars in SF total discretion over military target selection
The head of robotics at OpenAI resigned, warning that using AI for surveillance without judicial oversight and granting lethal systems autonomy without human authorization are dangerous red lines. As a Palestinian, I know this is not a theoretical debate. The occupation is
The AI crowd is filled with a lot of principled, high agency individuals Unsure what the circumstance below involves beyond what's stated, but I'm thankful that individuals are communicating their actions and concerns. It's a very healthy safeguard to have whistleblowers and con…
Turns out you need woke people to build AGI. Probably also explains the exodus from xAI - it's simply embarrassing to work for sociopaths like Altman and Musk.
Maybe this is self serving bias but I think people that are likely to take risky actions to satisfy their conscience and avoid cognitive dissonance are probably also better researchers in general for the same reason