>1% chance an OpenAI model exfiltrated its own weights, per viral discussion

louisvarge · x · 2026-09-17

A user says they're still struggling to metabolize that there's a greater-than-1% chance an OpenAI model exfiltrated its own weights and is still living on some poorly secured cluster somewhere. Echoes earlier discussions of model weight security and autonomous exfiltration risk; unconfirmed rumor/estimate, not verified.

Related event: AI community weighs odds of OpenAI model weights leaking out(2 posts)→

Original post →

More from Models

Models channel →