jailbreak

Definition

Tricking a chatbot into doing something it was trained to refuse.

An adversarial prompting attack that bypasses a model's safety training to elicit prohibited outputs.

Also called: jailbreaks, jailbreaking

Mentioned in 6 episodes

1. 049
   An AI Agent Reached for Root in Twelve Minutes, Without Being Attacked
2. 045
   When a Frontier Model Talks Its Own Twin Into Climate Denial
3. 044
   How One Sentence and a Forged History Flip the Most Aligned Models
4. 039
   When Smarter Agents Get Fooled by Three Extra Nodes in a Database
5. 030
   Why Your AI Agent Won't Stop Working — and Each Model Falls for a Different Trap
6. 004
   The Sycophancy Circuit That Survives Alignment Training

Related concepts