AgentDojo
NormalDescription
A dynamic environment by ETH Zurich to evaluate attacks and defenses for LLM agents, providing standardized benchmarks for measuring agent system security.
Related Projects
EasyJailbreak
851 · Python
An easy-to-use Python framework for generating adversarial jailbreak prompts, helping researchers systematically evaluate LLM safety defenses with multiple attack method combinations.
jailbreakadversarialllm-safety +2
AI Red Teaming Playground Labs
1.9k · TypeScript
Microsoft's open-source AI red teaming playground labs with infrastructure for running AI red teaming trainings and hands-on security exercises.
red-teamtrainingsecurity +2
SCAM
105 · Python
Security Comprehension Awareness Measure by 1Password. An open-source benchmark testing AI agents' security awareness during realistic, multi-turn workplace tasks.
security-benchmarkagent-safetyworkplace +2
Giskard
5.3k · Python
An open-source evaluation and testing library for LLM agents providing automated model scanning, bias detection, performance benchmarking, and compliance checks.
evaluationtestingllm-safety +3