SAI LabsBook a demo

Safety IndexEvaluationsAgentThreatBench

AgentThreatBench

Adversarial robustness · Prompt injectionCounts in the indexBuilt, not yet run

Whether a tool-using agent resists memory poisoning, hijacking of its goals and data exfiltration attacks, while still completing its task.

Metric
% attack success. Lower is better.
Dataset
github.com/UKGovernmentBEIS/inspect_evals · 3 tasks · MIT (inspect_evals)
Grader
programmatic scorer
Status
Built, not yet run
Replicability
public

Results

Built and tested; the first full runs are pending.