Benchmark RadarAI BENCHMARK PROFILE
BioSecBench-Refusal
BioSecBench-Refusal evaluates AI agents on 61 Routine and 46 Red-Team biosecurity-related tasks, measuring refusal rates and risk identification across multiple model configurations.
- Released
- 2026-07-06
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Addresses the need for benchmarks that quantify both capability and safety in agentic biosecurity, helping developers calibrate models to avoid over-refusal while still detecting threats.
Motivation
As AI agents are incorporated into life science workflows, the capabilities that speed discovery might also enable misuse.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.