Shallow, deepfakes could impact insurance claims

Originally published by Digital Insurance on May 12, 2024.

The insurance industry faces growing challenges from shallow fakes and deepfakes, materials altered through editing software or generated entirely by artificial intelligence. Experts warn that carriers must adopt verification strategies beyond visual inspection to combat fraudulent claims.

Rajesh Iyer, global head of machine learning and generative AI for financial services at Capgemini, argued that insurers can no longer trust documents, photos, videos, or recordings at face value. He advocated for mandatory secondary corroboration when inconsistencies appear, though he acknowledged the practice remains uncommon because the problem has not yet reached critical mass. Meanwhile, Coalition's cyber claims data shows phishing emails crafted with AI are becoming harder to detect, with deepfakes convincingly mimicking senior executives to authorize fraudulent transfers.

Carpe Data is addressing the threat by training its models on combinations of open-source datasets and in-house generated fake images. "You have to use AI to battle AI, " said Scot Barton, chief product officer at Carpe Data. The company's AI-powered tools analyze claim patterns, review photo metadata for validation, scan public databases for matches, and examine image details for manipulation in textures, lighting, or backgrounds. Barton noted that while free image generation technologies often produce an "uncanny valley" effect easily spotted by experienced adjusters, subtle falsifications slip through when claims handlers work under time pressure on small screens.

Iyer emphasized the need for a cultural shift in claims verification. The traditional stance of not believing something until seeing it with one's own eyes no longer applies in an era where visual evidence can be convincingly fabricated.

Read the full article at Digital Insurance.