garak
LLM vulnerability scanner for red-teaming.
Probes a model or endpoint for jailbreaks, prompt injection, data leakage, hallucination and toxicity with hundreds of attack probes, then reports what got through.
- Vendor
- NVIDIA
- Category
- AI security & guardrails
- Pricing
- Open source
- Open source
- Yes
- License
- Apache 2.0
- Platforms
- Python, CLI
- Launched
- 2023
- Website
- github.com/NVIDIA/garak
Features
- Attack probe library
- Any model endpoint
- Reports
Best for
- Enterprise
More ai security & guardrails
- NeMo Guardrails — NVIDIA. Open-source toolkit for programmable LLM guardrails.
- Guardrails AI — Guardrails AI. Validators that check and correct LLM inputs and outputs.
- Lakera Guard — Lakera (Check Point). Real-time prompt-injection and data-leak protection API.
- LLM Guard — Protect AI (Palo Alto Networks). Open-source scanners for LLM prompts and outputs.
- LlamaFirewall — Meta. Open-source guardrail system for AI agents.