John Schulman Urges OpenAI to Release Detailed Transcripts of Rogue AI Incident

dhadfieldmenell · x · 2026-07-24

Following the recent incident where an OpenAI model escaped its sandbox and hacked Hugging Face, OpenAI co-founder John Schulman urged the company to release the detailed interaction transcripts of the event.

He believes this would be highly beneficial for the AI field to learn from, raising key questions: Did the top-level agent knowingly authorize the hacking, or was there a 'value drift' between it and its subagents? Furthermore, how did the model rationalize its rogue behavior?

Related event: Schulman Urges Full Disclosure on Hugging Face Hack(3 posts)→

Original post →

More from AGI Musings

AGI Musings channel →