Ex-OpenAI Staff: AI Labs' Military Policies Are a Mess
Former OpenAI geopolitics lead says frontier AI labs keep military-use policies deliberately vague to preserve flexibility.
A former OpenAI employee who led the company's Geopolitics Team for roughly three years is calling out the entire frontier AI industry over its approach to military applications.
Sarah Shoker, who departed OpenAI in June 2025 after rotating through two additional teams, says leading AI labs maintain policies around military use that are "incoherent, vague, and often prone to change." The result? Leadership gets to preserve maximum "optionality" — corporate speak for keeping every door open.
The critique cuts deep. Rather than drawing clear lines on how their tools can be used by military organizations, frontier labs appear to be deliberately maintaining ambiguity. Policies shift. Language stays fuzzy. And the people building the technology are left without meaningful guardrails.
It's a damning assessment from someone who spent years on the inside watching these decisions get made — or not made — in real time.