cyber-evals
1 resources across 1 kinds
References
- Open ↗
Primary research from the UK AI Security Institute (DSIT): frontier-model cyber-capability evaluations, safeguard and jailbreak testing, and incident reports from their own agent testing.
1 resources across 1 kinds
Primary research from the UK AI Security Institute (DSIT): frontier-model cyber-capability evaluations, safeguard and jailbreak testing, and incident reports from their own agent testing.