Fraud is one of the insurance industry’s most persistent challenges, and the landscape is now further complicated by the adoption and use of artificial intelligence (AI). Fraud perpetrators are not the only adopters of AI, though. Insurers and plaintiff’s attorneys are also leveraging this emerging technology.
Industry research reflects this rapidly changing landscape. A CLM Magazine article covering the 2024 Anti-Fraud Technology Benchmarking Report by the Association of Certified Fraud Examiners (ACFE) and SAS noted that 83% of organizations expected to implement generative AI (genAI) as part of their anti-fraud programs, and the use of AI and machine learning in anti-fraud programs was expected to nearly triple by 2026.
Matthew Smith, executive director of the Global Insurance Fraud Summit, Inc. and president of Insurance Law Services, Inc., previously wrote in a CLM Magazine article, “Advances in AI, particularly in the fields of deep learning and neural networks, now enable the creation of highly realistic facial and voice simulations. Technologies such as deepfake algorithms can generate video and audio that closely mimic a real person’s appearance and voice.” Although highly sophisticated, these technologies can often be identified using specialized detection tools.
CLM approached its membership to share some of their experiences with AI-generated fraud and how they combat it in their own work.
AI In Action
Sarah Thomas, managing partner, Jones Jones LLC, shared her experience from the workers’ compensation sector, in which her firm is not yet seeing an influx of entirely AI-generated claims. Rather, “What we’re seeing are concerns about AI being used to support or embellish aspects of existing claims.” For example, she explains, “There have been discussions on the record about whether AI could be used to generate or assist with medical documentation, or to help create look-for-work reports. In New York, claimants who are partially disabled are often required to document their job search efforts, and AI has the potential to make those reports appear more complete or polished than the claimant’s actual efforts. Whether that’s occurring in a particular case is always fact-specific, but it’s certainly an issue practitioners are beginning to think about.”
Christopher Del Bove, partner, Callahan & Fusco, LLC, states that his firm has experienced “a few cases where plaintiffs have disclosed photos showing severe damage to the rear of their vehicles, which directly conflicted with photos in our client's possession from our own driver at the scene showing no damage.” Del Bove provides an example of one case in which “plaintiff made the mistake of filing an amended complaint, and we seized upon this and immediately filed a new answer with fraud defenses and a counterclaim, along with a notice to admit to ‘authenticate’ the photos that were disclosed. We strongly suspected that plaintiff's photos enhanced the damage to his vehicle with AI.”
Combating AI-Generated Fraud
“Carriers and third-party administrators (TPAs) are investing heavily in AI-driven fraud detection, document analytics, and data pattern recognition,” observes Thomas, “but the challenge is that genAI is improving at an incredible pace…. Ultimately, technology is only one piece of the puzzle. Experienced adjusters, investigators, attorneys, and medical professionals still play a critical role in identifying inconsistencies and asking the right questions.”
Del Bove notes that Callahan & Fusco’s in-house file management system has an approved AI tool that can assist in detecting AI-generated content. Often, he says, “one of the most effective tools in combating AI is to revert to trusted tools such as contacting claimants and verifying their stories. More carriers and TPAs are using AI to detect and map referral patterns.”
Challenges of Investigating AI-Generated Fraud
“One of the biggest challenges is proving an absence of something,” says Del Bove. “For example, proving that a photo is AI can be difficult, as there is always going to be a presumption that plaintiff's proofs are sincere.” Another challenge he experiences is law firms and carriers using commercially available and open-source AI, rather than trustworthy sources. “Our firm and clients have been systematically introducing AI with trusted partners that are aligned with our safety, security, and corporate responsibility concerns.”
Thomas notes that AI-generated materials often do not immediately look fraudulent. Experienced claims professionals, she explains, are often able to recognize inconsistencies in documentation, claimant language, or timelines. “From there, the investigation becomes much more traditional. You rely on testimony, cross-examination, medical records, employment records, metadata where available, and corroborating evidence.” She emphasizes that, although AI can change how questionable evidence is created, proving fraud “still comes down to good investigative work and testing credibility.”
Consequences of Widespread AI Adoption by Organized Criminal Fraud Rings
“If organized fraud rings begin using AI at scale, I think we’d see fraud become more sophisticated rather than simply more common,” Thomas says. “AI could allow bad actors to create convincing supporting documentation, coordinate fraudulent activity more efficiently, or tailor submissions to appear legitimate. That means the claims industry will have to respond in kind—combining advanced technology with experienced human judgment. I don’t think AI replaces investigators, adjusters, or attorneys. If anything, it makes their expertise even more valuable because someone still has to evaluate credibility, identify inconsistencies, and distinguish authentic claims from manufactured ones.”
Del Bove adds that “adoption of AI by fraud rings could lead to increased costs and premium hikes, as well as longer cycle times to resolve cases, as proving fraud often requires extensive discovery. Further adoption of AI has led to an ‘arms race’ in terms of firms, carriers, and TPAs having to invest in new technologies.”