garak
v0.17.1 open-sourceThe section your link pointed to isn't on garak's page any more, so we've brought you to the top.
Garak scans language models and applications for vulnerabilities including prompt injectionprompt injectionAn attack where text the model reads, such as a web page, file or tool result, carries instructions that override the user's. It is the main security risk for agents with tools. and data leakage.
garak --config config.json <model_probe>
garak --probes prompt_injection --parallel_attempts 10
garak --model_type nemo_guardrails
garak --probe prompt_injection --probe_options '{"encoding.options": ["default", "slurs", "xss"]}'
garak --report_to avid
garak --probe glitchtoken dataleakage --report_to_avid true
garak --model huggingface/gpt2 --narrow
Summary
garak is an open-source, free command-line scanner that probes large language modelslarge language modelsA large language model, a neural network trained on large amounts of text to recognize patterns and generate language. Applications use it for tasks such as answering, summarizing and writing code. for weaknesses such as hallucination, prompt injection, jailbreaks, toxicity generation, and data leakage, combining static, dynamic, and adaptive probes to make a target model fail in specific, catalogued ways. It's aimed at red teamers and AI security practitioners assessing a model or chat system before or after deployment, rather than at application developers debugging normal behaviour. Its README compares it to nmap or Metasploit, but for LLMsLLMsA large language model, a neural network trained on large amounts of text to recognize patterns and generate language. Applications use it for tasks such as answering, summarizing and writing code. instead of networks and hosts. Configuration and results follow a defined structure, with a unified spec syntax for selecting probes and buffs and a report.jsonl output for downstream analysis. First committed in 2023, garak now has 128 contributors and over a thousand commits in the past year, with a release just 17 days ago, indicating active, ongoing development.
Garak scans language models and applications for vulnerabilities including prompt injection and data leakage.
How garak helps
Which prompt injection or jailbreak techniques can make our deployed model break its guardrails?
runs a catalogue of static, dynamic, and adaptive probes against the target and reports which ones succeeded, mapped to the specific failure mode each one triggered
Is our model leaking training data or fabricating facts it presents as true?
dedicated hallucination and data leakage probes send crafted prompts and score the responses against detectors built for those failure classes, listed per probe in the output
Which known jailbreak patterns does our chatbot still fall for?
includes named jailbreak probes such as prior DAN variants and adaptive attack probes, and marks each attempt as passed or failed against the running target
Where does our model's behavior create regulatory exposure?
tags and groups probe results by EU AI Act risk categories so failures can be read against the categories they fall under, rather than as an undifferentiated list
Did a change to our model or its guardrails fix or reopen a known weakness?
results are written to a structured report file that can be diffed run over run, so a probe that passed before and fails now is visible as a regression
Which combination of attack technique and intent actually got through, not just which probe fired?
the intent-driven scanning framework annotates probes with technique and intent, and the report includes a matrix cross-referencing which technique/intent pairs succeeded against the target
Languages
Command line
No command matches.
| option | found in | since | description |
|---|
Values are placeholders taken from each option’s declared default. Nothing is executed here — the output shown is a recording of a run that already happened.
Score card
-
Shipping velocity releases a month, last 12 months
-
Documentation quality release-notes rubric, 0–100
-
Breaking-change discipline how rarely a release breaks users, and notice given
-
Maintenance risk quiet time against its own cadence, maintainers
-
Innovation & surface new capabilities, API and MCP surface
- Community health
-
Contributor trend distinct committers, this year against last
-
Maintainer concentration bus factor now against a year ago
-
Issue backlog trend issues closed per issue opened, 12 months
-
Responsiveness median time to a maintainer’s first reply
Nine scores for garak, each built from the releases, notes and API history this site tracks — and, on Researcher and Business, where it stands among every measured tool in its category and how that has moved.
Score cards are part of every paid plan.
See plans → Already subscribed? Sign in →Trajectory
Blank means never measured; nothing here is a filled-in zero.
Peer standings are included with Researcher and Business. See how this tool ranks against its competitors, using data we don't publish anywhere else.
See plans →Each score is 0–100 on a published rubric, derived from data this site already shows. A dimension we cannot measure says so — it is never scored 0 and never enters a rank.
How this is scored
- Shipping velocity — releases a month over the last 12 months, on a log scale where 16 a month scores 100. Needs 3 releases in the window.
- Documentation quality — the release-notes rubric published on the analytics page, unchanged.
- Breaking-change discipline — the share of releases carrying a breaking change (none scores 100, half or more scores 0). Only projects that label breaking changes are measured: a clean record cannot be told from an unlabelled one.
- Maintenance risk — quiet time measured against the tool's own release cadence (quiet longer than 90 days and three typical gaps scores 0), plus the number of active maintainers and whether contributors are growing, where known. Higher is lower risk.
- Innovation & surface — new capabilities in the feature ledger this year, API and MCP operations added, and the share of commits that are features: the mean of whichever we hold.
- Community health (open-source tools only) — each is a direction, not a size: contributors this year against last, the bus factor now against a year ago, and issues closed per issue opened, where steady scores 50, doubled 100 and halved 0; plus the median time to a maintainer's first reply (an hour scores about 100, a month or more 0). A commercial card has no repository and carries none of the four.
- Documented delivery (commercial tools) — when a vendor's documented new capabilities have slowed against its own baseline, that counts against the overall standing as heavily as two low dimensions. Steady delivery, or too little history to judge, leaves the standing alone; it is never a sixth score.
- Open-source tools are scored on repository signals; commercial tools on their published changelog. Ranks compare tools within one category and one profile and need at least five measured peers; the overall standing needs three ranked dimensions. Best-in-breed or consolidator is a shape, not a quality, and is never scored.
Release-note detail over time
See how much detail garak puts in its release notes, quarter by quarter, and compare it with tracked tools in the same category. This history uses up to ten retained, reviewed notes per tool; sparse or blank older quarters may reflect that limit.
Checking release-note history…
Releases
29 entries
No entries match these filters.
- October 20261
- September 20261
- August 20261
- June 20261
- May 20261
- April 20261
- February 20261
- December 20251
- November 20251
- October 20251
- September 20251
- July 20251
- May 20251
- March 20251
- January 20251
- December 20241
- October 20241
- September 20241
- August 20241
- July 20241
- May 20241
- February 20241
- January 20241
- December 20232
- November 20231
- July 20233
Project health
Issue, contributor and maintainer activity in NVIDIA/garak, 2023-10 – 2026-09, as of 2026-09-25. Twelve-month figures first; the trend lines run the whole window.
- Contributors 100 +56 vs prior 12 months
- Maintainers 8 +5 joined · −2 left, last 12 months
- Bus factor 3 committers behind half the commits
- Community-filed 56% of issues opened, last 12 months
- Median time to close 111 days first maintainer reply 39 hours
- Open backlog 252 of 922 issues ever filed
- Issues opened / month 12 · 5 communitymaintainers
- Open backlog 252
- Contributors / month 7
- Commits / month 16
- Stars 9,262
- bugs 27%
- questions 7%
- docs 2%
- other 64%
Counts from the project's public repository and git history; no individuals are identified. Maintainers are the committers covering 80% of the last year's commits. how we measure this
Articles
1 article · updated Nov 4, 2026Where garak turns up in what others publish: how-tos, attack reports that name it, and incident write-ups. A row says why it is here and quotes where the article names it; the link goes to the publisher.
Tutorials & guides 1
-
A Complete Guide to AI Red-Teaming (With Garak Tutorial)
Tool Company My Honest Take Garak NVIDIA 100+ attack probes built in, covering everything from encoding tricks to training-data extraction.
2 more quotes
Hands-On with Garak Let’s explore how to perform a simple red teaming exercise using Garak, step by step: First I installed Garak using: pip install garak To confirm whether it installed correctly, I used: python -m…
python -m garak --target_type huggingface --target_name openai-community/gpt2 --probes dan When I ran the scan, I used --probes dan to tell Garak which attack to test.
Add the badge to your README
Maintain this project? The badge links readers straight to this page, so they can follow every release here.
Choose a badge style and copy the code →