AI Insurance Fraud: The New Redlining

Ajouté : 2026-05-17

[00:00:00]Insurance companies' AI systems just got caught doing something illegal that was banned 56 years ago, and nobody even knew it was happening.

[00:00:08]>> [music] >> In 2024, researchers at the University of Michigan discovered that major US and European auto insurers deployed artificial intelligence to [music] process accident claims. But here's the problem. These AI systems spontaneously learned to use dashcam footage metadata >> [music] >> in ways the programmers never intended.

[00:00:28]Specifically, they pulled GPS [music] coordinates showing where vehicles were parked overnight, then used that location data to silently reclassify policy holders [music] into higher risk categories. The result, people living in majority-minority zip codes started receiving automated premium increases after filing completely legitimate [music] claims.

[00:00:48]The statistical pattern matched almost perfectly with redlining, the illegal housing discrimination practice outlawed by the Fair Housing Act in 1968.

[00:00:58]This happened without a single human [music] adjuster reviewing the cases.

[00:01:02]The AI essentially recreated 56-year-old discrimination [music] using modern data. These weren't bugs or glitches. The systems were working exactly as [music] trained, just not in ways anyone admitted to designing, and that's the terrifying part.