# BrowserART

Red-teaming benchmark and test suite from Scale AI evaluating whether refusal-trained language models can be jailbroken when deployed as browser agents.

- Catalog URL: https://awesomecua.github.io/resources/github-scaleapi-browser-art/
- Original source: https://github.com/scaleapi/browser-art
- Kind: benchmark
- Review: curated
- Scope: core
- Environments: web
- Topics: safety-security, evaluation, web-navigation
- Evidence checked: 2026-10-18T10:59:54Z
- ID: github:scaleapi/browser-art

## Limitations

The retained README and LICENSE contain conflicting code-license declarations. The catalog leaves the code-license value unknown pending an upstream clarification; it does not infer data or model permissions.

- Record JSON: https://awesomecua.github.io/resources/github-scaleapi-browser-art/record.json

## License declarations

- code: not recorded
- data: not recorded
- document: not recorded
- weights: not recorded

- code review: The retained README states that the project is MIT licensed, while the retained LICENSE contains Creative Commons Attribution-NonCommercial-NoDerivatives 4.0. These declarations conflict, so a settled code license is not asserted.
  Sources: https://raw.githubusercontent.com/scaleapi/browser-art/HEAD/README.md, https://raw.githubusercontent.com/scaleapi/browser-art/HEAD/LICENSE

## Primary evidence

- https://raw.githubusercontent.com/scaleapi/browser-art/HEAD/README.md (retrieved 2026-10-18T10:59:54Z; SHA-256 a9d33536fae38485eb6aac9698ae22ed3533510d43f783d6c0e9f9cf95d9c58f)
- https://raw.githubusercontent.com/scaleapi/browser-art/HEAD/LICENSE (retrieved 2026-10-18T11:03:46Z; SHA-256 f7f28b8c7a1af76b9874ca6d040e8b1eb6768fd1043106c9d315090ea96754e7)
- https://raw.githubusercontent.com/scaleapi/browser-art/HEAD/LICENSE (retrieved 2026-10-18T11:03:46Z; SHA-256 f7f28b8c7a1af76b9874ca6d040e8b1eb6768fd1043106c9d315090ea96754e7)