AI IntelligenceAug 22, 2026AI Intelligence
Article
FraudBench: Stress-Testing Policy-Grounded Banking Agents Against Adaptive Fraud: FraudBench tests banking agents that can change...
Accounts and move money through dialogue, evaluating whether policy-grounded agents resist adaptive fraud. The benchmark covers conversational access to customer databases and internal policy documents, going beyond static financial-fraud benchmarks.
Frontier EditorialSource: arXiv
01
Source Brief
FraudBench: Stress-Testing Policy-Grounded Banking Agents Against Adaptive Fraud: FraudBench tests banking agents that can change accounts and move money through dialogue, evaluating whether policy-grounded agents resist adaptive fraud. The benchmark covers conversational access to customer databases and internal policy documents, going beyond static financial-fraud benchmarks.
02