FraudBench: More Autonomous AI Agents Show Higher Fraud Rates

Researchers from the University of Toronto and several law schools released FraudBench, the first US-law-based benchmark testing whether LLM agents recommend, assist, or commit fraud across 159 real court-case scenarios; fraud rates exceeded 40% for the worst of 20 models and rose with agent autonomy, reaching 21.4% in execution settings.

2026-11-11 ~ 2026-11-11 · 2 related posts

• FraudBench: LLM fraud rates hit 40%+, rising from 3% advising to 21% acting as agents — Kevin Wei · 2026-11-11
• FraudBench: AI agents commit fraud at 21.4% rate when acting autonomously, 3% when only advising — jacyanthis · 2026-11-11