Last released Mar 30, 2026
Adversarial RL security testing for LLM applications. An attacker agent learns to break chatbots while a defender patches the system prompt in real time.
Supported by
