Skip to content
  1. Home
  2. AI tools
  3. garak

garak

LLM vulnerability scanner that probes models for known failure modes

What it is

A red-teaming kit from NVIDIA that tests whether an LLM or LLM-backed application can be made to fail. It runs static, dynamic and adaptive probes for prompt injection, data leakage, hallucination, toxicity and other weaknesses, similar in spirit to a network scanner. For security teams assessing a model before deployment.

Good for

  • Scan a model for prompt injection weaknesses
  • Check an LLM app for data leakage
  • Produce a pre-deployment risk report

Install or run

python -m pip install -U garak

Command as given in the project README. Read the repository before you run anything on your machine.

What you need

Python; access to the model or endpoint under test

Worth knowingFull probe runs make many model calls and can be slow and costly; use it only on systems you are authorised to test.

Sources

The summary is ours, written from the README. Spotted an error? See how we work.

More for security & quality

See the topic →
  • Security Guidance plugin

    NousResearch/hermes-agent

    Warns when written code contains known-dangerous patterns

    ToolGitHub stars: 263k
  • security-guidance

    anthropics/claude-plugins-official

    Security warnings on edits plus LLM diff review

    ToolGitHub stars: 47k
  • Langfuse

    langfuse/langfuse

    Open-source tracing, evals and cost tracking for LLM applications

    ToolGitHub stars: 37k
  • Promptfoo

    promptfoo/promptfoo

    CLI for evaluating and red-teaming prompts, agents and RAG apps

    ToolGitHub stars: 27k