Japanese insurers build detectors for AI-faked claim photos

What the insurers are doing

As 政経フォーカス on MSN reported, citing Mainichi Shimbun, Japan's property and casualty insurers are stepping up defences against claims backed by AI-made pictures. These are the companies that cover cars, homes and belongings.

Mitsui Sumitomo Insurance worked with group companies, including Aioi Nissay Dowa, to build a system that tells whether an evidence photo has been doctored by AI. It has been running since August.

What the detector found

In September, the first full month after launch, several thousand people filed claims for fire, injury and similar losses. A small number of them submitted photos the system judged to be AI-generated. That number was higher than in the previous month.

The company says it fed data from past criminal investigations into the detector to make it more accurate. The report does not explain how the tool works beyond that.

From gangs to everyday cheating

Mitsui Sumitomo once saw organised fraud rings file a steady stream of false claims. The groups recruited part-timers, signed them up for policies online, then claimed with edited photos. A dedicated countermeasure team and closer manual review brought suspected cases of that kind down to zero.

Now, with image generation improving, fake-photo claims are rising again. The new cases tend to be small and individual: a faked car scratch, or damage to a laptop, phone or small appliance, usually worth tens of thousands of yen. Claims that size once seemed not worth the effort of faking.

The wider industry response

Other big insurers are moving too. Tokio Marine & Nichido has an AI-based fraud detection system, and Sompo Japan has set up a specialist department. Mitsui Sumitomo's spokesperson said the firm will take a firm line on AI-enabled fraud as it spreads across the industry.