garak is an open-source command-line LLM vulnerability scanner and red-teaming assessment kit. It probes large language models for failures such as hallucination, data leakage, prompt injection, misinformation, toxicity, jailbreaks, and other weaknesses. garak supports many backends (Hugging Face, OpenAI, Replicate, AWS Bedrock, gguf/llama.cpp, etc.), provides a wide range of probes and detectors, and produces structured logs and JSONL reports. Licensed under Apache-2.0.
