AI Red Teaming Playground Labs

Vigil is an LLM security detection tool that identifies prompt injections, jailbreaks, and other potentially risky LLM inputs through multi-dimensional analysis for real-time safety protection.

An easy-to-use Python framework for generating adversarial jailbreak prompts, helping researchers systematically evaluate LLM safety defenses with multiple attack method combinations.

A dynamic environment by ETH Zurich to evaluate attacks and defenses for LLM agents, providing standardized benchmarks for measuring agent system security.

An open-source benchmark for prompt injection attacks and defenses in LLMs, systematically evaluating the effectiveness of different attack strategies and defense mechanisms.