Artificial intelligence has made it easier to fabricate images for fraudulent claims, requiring insurers to scrutinize evidence from policyholders more closely than ever.
It's an issue that the industry is deeply aware of. In a recent study by Verisk, 98% of insurers said
In many cases, the metadata can tell a story that the eyes can't see, according to Danielle Lutkus, principal consultant at Capco. Insurance companies can use this data to detect patterns and empower their underwriting and fraud teams to better respond to AI-driven fraud, Lutkus shared with Digital Insurance.
Responses have been lightly edited for clarity.
How are claims, underwriting and fraud teams responding to AI-created photos?
Carriers are approaching AI-created photos in a few ways, often layering different techniques to ensure coverage against fraud. The important first step is to treat all new images as unverified until they are vetted through a review process. These review processes may include metadata analysis, reverse image searches, human-in-the-loop review of images for anomalies and can lead to requests for onsite inspection by an adjuster.
In an effort to get ahead of fraud without slowing down the process, some carriers are issuing their own apps that embed guided prompts to capture very specific images of accidents or damage directly within the app. This allows them to view the images in real time and capture the relevant metadata to validate the images accuracy.
How does metadata forensics and anomaly detection work?
In the review of images, it's important to look beyond just the visuals. Metadata forensics examines the information stored with the image that outlines size, when it was created and changes that have occurred to uncover suspicious attributes. Anomaly detection can spot trends on a series of images to see if the creation dates align, the location data is in sync, and there are no unexpected file-modification dates.
Carriers should establish a repeatable process that outlines a scorecard of metadata attributes so that when an image falls outside of the baseline, additional validation processes can be triggered to review the anomaly like manual review and claims adjuster onsite review.
What kind of impact does this have on documentation, client expectations and claims speed?
Carriers' implementation of new processes for advanced image analysis, metadata validation and anomaly detection to strengthen confidence in submitted evidence will impact both policyholders and internal processes.
Policyholders will need to adjust their expectations around claims processing, documentation requirements and speed to claim. They may encounter additional verification steps, requests for original files, or enhanced review procedures. Advisors should accordingly prepare clients for a future in which stronger evidence validation becomes a standard component of responsible risk management.
The rapid evolution of AI and ongoing digitization is speeding up insurers' shift to more digital-driven processes. Policyholders are now benefitting from carriers developing apps and more digitally driven processes that can make the process easier to follow and status more transparent for the policyholder.
What advice do you have for insurance companies that are facing issues like this?
Insurers need to take a 360-degree approach to examining their fraud procedures and determining areas of weakness that are being targeted by AI-driven fraud. When weaknesses are identified, carriers should take a balanced approach to implementing tactics at the people, process and technology level.
Implementing very heavy technology-driven fraud detection measures can make claims or new business processes too arduous for distributors or policy holders. Balancing technology approaches with setting documentation quality expectations with agents and investing in more transparent and user-friendly tools, can offset the weight of added processing time being solely at the carrier level.
Why does documentation need to be treated as a risk-management discipline and what does that mean?
Documentation provides another key data element needed when compiling all the information that carriers use to make informed risk management decisions. Documentation determines how new business, underwriting and claims are evaluated and scored to influence the outcomes of those processes. It's important that carriers are disciplined in the evaluation of risk to improve consistency, predictability and outcomes for their policyholders.
As an example, if a life insurance carrier underwrites policies based on fraudulent medical records, they may face losses due to early claims. AI document analysis can accelerate risk prevention by identifying, examining and escalating documents that may be considered fraudulent to an agent or claim examiner, which can protect the company from bad actors.
