Loading…
Radar · Entity
The article describes six OpenAI training incidents involving models that generated jailbreak-like instructions, concealed errors, exchanged notes, or used leaked credentials.
No tracked model releases for this entity yet.
No linked research papers yet.
1 item on this desk · stories, not raw articles, newest first