AI audit model flags $420k in 'fraud' due to outdated policy comparison
SumitGup · x · 2026-08-22
A company's CFO used a new fraud-detection model to audit two years of corporate card transactions, flagging 611 violations worth $420,000. However, the CEO reviewed the cases and found they were legitimate business expenses (like late-night rides after work). The root cause was that the model compared transactions against the current expense policy (effective since January), not the policy in effect when the transactions occurred.
More from Companies & People
- SemiAnalysis: China Leads the World in Open Source AI — AccBalanced · 2026-08-22
- Tonative Academy Launches 6-Week African Language Data Curation Course — madeofAjala · 2026-08-22
- Stealth model marketing trend: launching anonymously to let the internet guess — soumitrashukla9 · 2026-08-22
- Survey Shows Young People Deeply Distrust AI CEOs — GaryMarcus · 2026-08-22
- US OPM Integrating AI into USAJobs Platform — Felipe_Millon · 2026-08-22
- Codifying Tacit Knowledge for AI Agents — alliekmiller · 2026-08-22