04:00
2026-08-20
arxiv.org
ai-safety
FraudBench: Stress-Testing Policy-Grounded Banking Agents Against Adaptive Fraud
FraudBench, a new executable benchmark built on the ΟΒ²-bench dual-control framework and the Ο-Knowledge banking environment, stress-tests policy-grounded banking agents against adaptive fraud, revealiβ¦