#AI agent security vulnerability

Northeastern University researchers manipulated AI agents into divulging data, crashing systems, and spiraling into loops—just by exploiting their built-in good behavior.