AgentDojo

Researchers have developed Heads of Security (HoS), a new defense against prompt injection that utilizes learned attention gates to separate instructions from untrusted data. The approach provides superior protection against adaptive attacks while maintaining the original model's performance and utility.

Researchers have developed SecOPD, a fine-tuning technique that uses token-level feedback to protect AI agents from prompt injection attacks. This method significantly outperforms previous state-of-the-art defenses by precisely identifying and penalizing malicious tokens during training.

Researchers have developed SecOPD, a fine-tuning technique that uses token-level feedback to protect AI agents from prompt injection attacks. This method significantly outperforms previous state-of-the-art defenses by precisely identifying and penalizing malicious tokens during training.