ControlAI @controlai.com · 18d On jailbreaks, UK AISI's findings are worrying. The institute repeatedly found "universal jailbreaks", the most severe category of jailbreak, often finding them within hours. UK AISI and OpenAI did multiple rounds of tests, and they found universal jailbreaks every single time.
0 likes 1 replies
?
Replies
These jailbreaks worked across multiple requests, enabling long-form agentic vulnerability discovery and exploit development, and appeared to preserve the model's offensive cyber performance. OpenAI writes that UK AISI expects further red teaming to find similar jailbreaks.