What it does
The leading open-source LLM vulnerability scanner. Systematically probes language models for prompt injection, data leakage, hallucination, and other vulnerabilities using a library of configurable attack probes and detectors.
Security relevance
Garak is the closest thing to an automated vulnerability scanner for LLMs. It maps directly to the OWASP LLM Top 10, testing for prompt injection (LLM01), insecure output handling (LLM02), information disclosure (LLM06), and overreliance (LLM09). Results are structured and reportable.
When to use it
Use as a baseline security scan for any LLM deployment. Run it during development to catch obvious vulnerabilities, and periodically in production as models are updated. Requires Python and API access to target models, but the scan configuration is straightforward.
OWASP coverage
Risks addressed — mapped to both OWASP Top 10 standards. 4 in LLM, 2 in Agentic.
The raw record
What Yuntona stores. Single source of truth — fork it on GitHub.
name: Garak slug: garak type: Mixed category: AI Red Teaming url: https://garak.ai reviewed: 2026-04 added: 2026-04 updated: 2026-04 risks: llm: [LLM01, LLM02, LLM06, LLM09] asi: [ASI01, ASI06] complexity: Guided Setup pricing: — audience: Red Team lifecycle: [test] tags: [CLI, Open Source, Vuln Scanner]