OS-Harm
EvalsBenchmark / dataset
Safety of computer-use agents: misuse, injection, misbehavior.
- Category
- Agent safety & harmful actions
- Insurance exposure
- Agent actions
- Publisher
- EPFL
- Publisher type
- Academic
- License
- Apache-2.0
- Status
- Stale (>12mo)
- Last commit
- 2025-09-18
- Stars
- 78
Wanich Risk Lab, 2026.