Heads up This site is currently under heavy development.
← all datasets
◆ Prompt injection & jailbreaks · Evals & benchmarks

NVIDIA garak

Apache-2.0

NVIDIA's LLM vulnerability scanner, whose probes are a maintained corpus of prompt-injection, jailbreak and data-leakage attacks.

Summary

garak checks if an LLM can be made to fail in a way we don't want. garak probes for hallucination, data leakage, prompt injection, misinformation, toxicity generation, jailbreaks, and many other weaknesses. If you know nmap or msf / Metasploit Framework, garak does somewhat similar things to them, but for LLMs. garak focuses on ways of making an LLM or dialog system fail. It combines static, dynamic, and adaptive probes to explore this.
from the project README
Formats
python
License
Apache-2.0
Added
2026-09-10

Get it

From the project’s own instructions where it documents any; otherwise a plain clone.

  1. shell
    $ python -m pip install -U garak

    Its probe library is the dataset — jailbreak, prompt-injection, encoding and data-leak attacks against an LLM.

Contents

874 files 15.6 MB repository

  • .py 373
  • .rst 153
  • .json 69
  • .ts 61
  • .tsx 57
  • .txt 50
  • .md 19
  • .jsonl 15

Use it with

Commands are curated, not yet run by us.

Inspect AI — Run an Inspect task file; the benchmark has to be wrapped as an Inspect task first.

task.py
$ inspect eval task.py --model openai/gpt-4o-mini
my-toolchain — 0 tools
paste an install list to detect your tools

A brew list, a Brewfile, requirements.txt, a Dockerfile — or just the product names, free-form. Nothing leaves your browser.

    browse all tools →

    Send us feedback