AI makes fraudulent claims plausible

Claims Journal reported on May 15 that generative AI tools are making fraudulent insurance claims look professionally assembled, with fake photos, invoices, medical records and identities now able to pass an initial review. The article, by Kayla McCallum, said the change is not that insurance fraud is new, but that the effort needed to make a false claim look legitimate has fallen sharply. That shift is forcing claims investigators to spend less time looking for crude red flags and more time checking whether evidence is authentic. Industry reports and legal guidance published this year describe the same pressure on carriers, claims teams and special investigation units. ### How are false claims being made to look real? Claims Journal said fraudsters can now use widely available AI tools to create photographs, invoices, medical records and “entire identities” that can survive an initial screening. The publication described the emerging problem as a “synthetic claim” — a file assembled from believable but fabricated parts, sometimes mixed with real data and attached to a fictitious person. (claimsjournal.com) Debevoise & Plimpton wrote on January 20 that insurers are especially exposed because many claims workflows already rely on submitted images rather than original objects or documents. The law firm said newer generative AI models can fabricate photorealistic images of vehicle or property damage, creating the risk of nonexistent or exaggerated losses. Guidewire said on January 16 that claims teams are encountering several forms of manipulated media: deepfakes, simple edited images and fully synthetic content, including fabricated voices, videos and businesses. (claimsjournal.com) The company said traditional review methods and legacy fraud models were not built to detect those manipulations. ### What has changed for investigators reviewing a file? (debevoise.com) Kayla McCallum wrote that investigators historically could start from the premise that most of a claim file was genuine, with one suspicious document or inconsistency standing out. That assumption “can no longer be a safe starting point,” she wrote, because multiple parts of a submission may now be fabricated together and made to reinforce one another. (guidewire.com) Claims Journal said that changes the first task in an investigation. Instead of asking whether a file contains an obvious anomaly, adjusters and SIU teams now need to establish whether the underlying material is original, consistent and independently corroborated. The article also warned that carriers cannot simply deny claims because something appears unusual, because that creates bad-faith exposure if a legitimate claimant is rejected on suspicion alone. (claimsjournal.com) ### What evidence are carriers being told to ask for? Claims Journal said every investigation should begin with original files rather than screenshots or forwarded attachments. Native files can preserve creation dates, device information, GPS coordinates and edit history that may not survive once a file is copied or compressed, the article said. Debevoise & Plimpton recommended several similar checks, including asking for multiple photos from different angles and distances, requiring reference objects for scale, requesting original documents by mail and using real-time remote video inspections in which an adjuster directs what the claimant shows on camera. (claimsjournal.com) The firm also said insurers may need more on-site visits in some cases. Guidewire said authenticity controls should be embedded in claims operations as automation increases. The company said insurers need tools that can verify where digital content came from, whether it was altered and how confident reviewers can be in its authenticity. ### How widespread do insurers say the problem is? Verisk said in its 2026 State of Insurance Fraud Report that 98% of insurers believe AI editing tools are fueling digital fraud. (debevoise.com) The company said only 32% of insurers surveyed felt “very confident” detecting deepfakes. The same Verisk study said 55% of Gen Z respondents would consider editing a claim photo or document, while 69% of consumers said fraud would raise premiums for all policyholders. (guidewire.com) Verisk said the findings were based on surveys of 1,000 U.S. consumers and 300 insurance claims professionals. Guidewire, citing Shift Technologies, said 20% to 30% of insurance claims may now include altered images, fabricated documents or synthetic medical reports. (verisk.com) That figure was cited in a vendor blog rather than a regulatory filing or public enforcement dataset, but it shows the scale insurers and technology providers are using to describe the threat. ### What does this mean for SIU teams and claim referrals? (verisk.com) Claims Journal said SIU and claims teams need to focus on corroboration and referral quality rather than simply generating more alerts. The article argued that a larger volume of suspicious signals is less useful if investigators cannot tell which files contain authentic evidence and which are built from synthetic materials. Verisk’s report and the legal guidance from Debevoise point in the same direction: carriers are being pushed toward intake controls, source verification and better documentation of why a file was escalated. (guidewire.com) Those steps matter because a manipulated claim may require a fraud referral, partial payment decision or outright denial, depending on what can be proved and what the policy language allows. (claimsjournal.com) Verisk’s 2026 State of Insurance Fraud Report remains available from the company, and Claims Journal’s May 15 article identifies document review, native-file collection and corroboration as immediate steps for claims and SIU teams. (claimsjournal.com) (verisk.com)