garak is an LLM vulnerability assessment toolkit developed by NVIDIA designed to evaluate whether language models can be made to fail in unwanted ways. It combines static, dynamic, and adaptive probes to explore failure modes such as prompt injection, data leakage, hallucination, toxicity generation, and jailbreaks. Modeled after security tools like nmap and Metasploit Framework, it serves as a red-teaming toolkit for identifying security and reliability weaknesses in deployed LLM systems.