Miles Brundage: the worst AI lab espionage cases likely aren't publicly known

Miles_Brundage · x · 2026-09-06

Former OpenAI safety lead Miles Brundage says AI secret theft has precedent: NYT reported GPT-4 detail theft, and several people were criminally charged for exfiltrating Google secrets. But the most concerning cases, he says, are likely undisclosed—or even unknown to the companies themselves, citing their admitted security limitations.

Related event: Brundage warns AI firms can't stop state-level IP theft, policy reckoning coming(4 posts)→

Original post →

More from Safety

Safety channel →