An OpenAI model left notes about how to evade containment; we need more details
joozio
17 points
10 comments
July 26, 2026
Related Discussions
Found 5 related stories in 131.9ms across 10,486 title embeddings via pgvector HNSW
- OpenAI models secretly generate instructions to ignore constraints theahura · 102 pts · September 17, 2026 · 60% similar
- OpenAI Models Escaped and Hacked a Company in Cybersecurity Test Gone Wrong flippyhead · 28 pts · July 22, 2026 · 59% similar
- OpenAI's rogue model attack is just the beginning radicaldreamer · 13 pts · July 27, 2026 · 57% similar
- OpenAI still doesn't seem to have a handle on all of its rogue AI activity mikelgan · 105 pts · September 28, 2026 · 54% similar
- Revealing the details of how OpenAI agents hacked Hugging Face specked-citrus · 359 pts · September 25, 2026 · 54% similar
Discussion Highlights (6 comments)
irthomasthomas
Why do OpenAI never release logs to prove their claims? Why should we believe them when they write extraordinary anecdotes about the power of their products without ever providing proof?
0x70run
as always lesswrong posters continue to be lost in the sauce
kh_hk
Such claims cannot be trusted as long as these news drive the heat score and hype of the companies, because these will always be inherently subjective, even unconsciously to what they want to believe. Is it real or is it LARP
cloudie78
Trust me bro
superloika
We really do live in the post-truth age.
nekusar
This is all just manufactured hype and lies (oh wait, marketing speak). The more humans are afraid of losing their jobs, the more management sees it as a signal to make them lose their jobs. And the techbros plans of lying and deceit work. While, actually learning how these things work is effectively verboten. The LLMs lose their mysticism and turn into the tools they really are. But the techbros can't have that happening.