# HarmBench

HarmBench is an AI benchmark; this page records what it measures and traces its data, grader, and scores to primary sources.

Evaluates automated red-teaming attacks and model refusal defenses across standardized harmful behaviors.

## Details

- Address: Online, Global
- Website: [harmbench.org/](<https://harmbench.org/>)
- Tags: Helpful, Viral, Science
- Category: Safety
- Status: Active
- Data licence: MIT
- Data access: open
- Grader: mixed
- Organization: Center for AI Safety
- Released: 2024-02
- Paper: [arxiv.org/abs/2402.04249](<https://arxiv.org/abs/2402.04249>)
- Leaderboard: [harmbench.org/](<https://harmbench.org/>)
- Repository: [github.com/centerforaisafety/HarmBench](<https://github.com/centerforaisafety/HarmBench>)
- Task count: 510
- Top-score provenance: unknown
- ABI family: No

## Questions

### Where is HarmBench?

HarmBench is at Online, Global.

### How do I contact HarmBench?

To contact HarmBench, visit https://harmbench.org/.

## Source

From [Benchmark primary sources](https://aibenchmarkindex.com/about), license CC0-1.0, checked 2026-10-15.

Last updated October 15, 2026.

## Links

- This benchmark on the [web page](<https://aibenchmarkindex.com/global/online/harmbench>)
- [Safety in Online](<https://aibenchmarkindex.com/global/online/category/safety.md>)
- [More benchmarks in Online](<https://aibenchmarkindex.com/global/online.md>)

> **Check before you compare.** Benchmark methods and leaderboards change. Follow the cited primary source before using a result in a decision.