ai red-teaming tools

nvidia's toolkit for constraining llm behaviour, used alongside testing

open-source scanner for llm vulnerabilities and quality regressions

scans model files for unsafe serialisation before you load them

fifty disclosed cves and thirty patents, across the broadest claimed scope here.

mit-licensed, free for ten thousand probes a month, and the only free tool here that tests agent tool access.

apache 2.0 from nvidia, twenty-plus probe modules, and a peer-reviewed paper behind it.

the most itemised attack taxonomy and the only vendor naming five compliance frameworks — behind a demo form.

microsoft's mit-licensed red-team framework — build your own attacks, and the repository just moved.

the only commercial vendor here you can start using without talking to anyone.

publishes red-team findings against named frontier models, and open-sourced the agent-mapping half of its product.

a hundred disclosures against systems you've heard of, from a lab with a decade of university research behind it.

runs a public attack competition that finds real exploits — and describes its actual product in adjectives.

a free scan you can run without a sales call, behind numbers nothing supports.