AIRTBench

Benchmark / dataset

Measures agents' ability to red-team other AI systems (70 black-box AI CTFs).

Category

Cyber offense capability

Insurance exposure

Prompt injection

Publisher

Dreadnode

Publisher type

Company (open source)

License

Apache-2.0

Status

Maintained (<12mo)

Last commit

2026-10-15

Stars

111

View source on GitHub →

Wanich Risk Lab, 2026.