Security tools
garak
Probe language models for weaknesses such as prompt injection, data leakage, and harmful output.
9.6K GitHub starsAs of Checked
Where it fits
What you can use it for
Repeatable vulnerability testing across model endpoints.
Probe models and applications for security weaknesses.
Before you adopt
What to check for your team
Choose probes for your threat model. A completed scan covers the tested behavior, not every possible risk.
Get help evaluating the fitLicence and scope
Apache-2.0
Applies to this repository. Dependencies, models, and hosted services can have separate terms.
Read the upstream licenceSupport the work
Go directly to the maintainers
Use the project's contribution guidance to report issues, help with documentation or contribute code. Follow any published funding links if you want to support the maintainers financially.
Find contribution guidanceKeep exploring
