A primer on insurance fraud detection and the power of GenAI technology

Generative artificial intelligence (GenAI) has shaken up the world with its ability to create text, images and music bearing an uncanny resemblance to human-created materials. GenAI has also shaken up the insurance industry with its ability to both prevent and perpetuate insurance fraud at a speed previously unimaginable.  

This dual potential makes GenAI an important focus for insurers. To effectively combat modern fraudsters, insurance professionals should understand and leverage GenAI themselves. Consider this your GenAI insurance fraud primer.  

What GenAI is and how it’s different from conventional AI 

Let's first establish the differences between GenAI and traditional AI capabilities. One of GenAI’s primary distinctions is it generates net-new content that can mimic human creations.  

As an example, take this GenAI chatbot response when prompted to contrast conventional and generative AI: “Traditional AI focuses on tasks like classification, prediction or decision-making based on existing data, while Generative AI goes a step further by learning patterns from existing data to generate new and original content that closely resembles human-created content.” GenAI generated this original sentence after reviewing relevant information related to the prompt or question. On the other hand, a customer service chatbot regurgitates a whole sentence from a repository.  

So, GenAI’s ability to better comprehend context, mimic the human voice and do it all with impressive speed can improve productivity for anyone, no matter their goals. 

GenAI has taken fraudulent claims to the next level 

Insurance fraud schemes may be more sophisticated thanks to technology, but they’ve existed for centuries. One of the earliest known examples of insurance fraud dates to around the 4th century BC. (Yes, a type of insurance existed then.) To collect insurance money, a Greek merchant named Hegestratos intentionally sank his ship and claimed it was lost in a natural disaster. Luckily, ancient Greek insurers were savvy enough to find him out. 

In modern times, fraudsters don’t have to sink entire ships to make money. Let’s use an auto accident as an example. Before the digital age, a neck brace might suffice to feign an injury requiring compensation. As personal computers became commonplace, an individual might forge vehicle invoices or repair bills. Conventional AI algorithms made altering an array of documents easier — invoices, bills, claim applications, vehicle details, ownership history, etc. Scammers have always found ways to scam. 

So, it’s no surprise fraudsters see opportunity in GenAI, which is both highly adaptable and widely available. Some GenAI-generated fraudulent techniques we should expect to see more of include: 

1. Increasingly convincing fake documents: GenAI models have a deeper understanding of context, language patterns and grammar. This means GenAI-generated insurance policies, medical records, accident reports, identification papers and invoices can resemble materials made by humans with less effort by the fraudster. At the very least, they won’t resemble comically unbelievable phishing emails generated by conventional AI frameworks. 

2. Detailed synthetic identities: Creating fake names, addresses and employment histories may not be new, but GenAI does this with more coherence and consistency. For example, a GenAI-created fraudulent software engineer might have a computer science degree and a tech-centric work history. This level of believability and consistency is unlikely to raise red flags upon review. 

3. Realistic-looking doctored images and videos: As an example, fraudsters can create images of auto accidents or property damage that never occurred. With GenAI, a photo can depict an accident scene that looks authentic in terms of vehicle positioning, collision impact and damage distribution. GenAI can even account for shadows, reflections and proportions. 

While these approaches contain a convincing level of detail, fraudulent schemes can be even more elaborate. Commodity Futures Trading Commission (CFTC) Commissioner Kristin Johnson relayed one “Ocean’s Eleven-styled operation” in a February 2024 speech: A branch manager received a call from a director he spoke to frequently, instructing him to transfer $35 million for a pending acquisition. Emails confirmed the legitimacy of the request, so the branch manager obliged. However, fraudsters and some sophisticated voice-cloning technology were behind the whole affair. One takeaway to keep in mind: Celebrities aren’t the only victims of deep fakes.   

GenAI has taken insurance fraud detection to the next level too 

Thankfully, insurers have this same advanced technology at their disposal to combat fraudsters. GenAI, in tandem with human oversight, can just as innovatively help prevent fraud — and in comparatively less time than a manual or conventional AI approach.  

Here are some ways GenAI can assist the insurance industry: 

1. Anomaly detection and behavioral analytics: GenAI can analyze data from insurance claims and policyholder behavior to help identify unusual patterns, including more complex behavioral changes. For example, if an individual suddenly submits multiple high-value claims across different policies within a short period, AI algorithms can flag these activities as anomalies for further investigation.    

2. Document verification: It can learn from a wide range of examples and variations, allowing it to detect manipulated images, recognize inconsistencies in document metadata and identify subtle discrepancies that may indicate fraud. 

3. Extracting key fraud characteristics from unstructured data: Generative AI can process unstructured data that’s not in a specific format, like a spreadsheet, and extract more complex fraud characteristics. For instance, it can analyze adjuster notes to help identify patterns of exaggerated or inconsistent descriptions of events or analyze images to detect signs of staged accidents or property damage that do not align with the reported incident. It can basically detect hidden fraud patterns that’s difficult for both the human eye and conventional AI to see. 

By leveraging these capabilities, insurers can help defend against sophisticated fraud tactics and enhance their detection and prevention strategies, enabling a more secure and reliable insurance system.  

How insurers can use AI responsibly 

The Responsible AI movement — AI practices to preserve value and engender safety and trust — has grown alongside the increasing use of GenAI, as it should. All insurers should aim to leverage AI responsibly, and third-party AI tools can help supply these frameworks, so GenAI use aligns with regulatory requirements. 

Risk Detect, a PwC product, is one such GenAI-enabled tool to help insurance carriers combat fraud, corruption and compliance issues. It leverages GenAI to analyze vast amounts of data and identify patterns and anomalies that can indicate fraudulent or suspicious activity.  

Battling insurance fraud beyond AI 

When it comes to insurance fraud, you can’t afford to be flat-footed. Both the technology and the schemes will likely improve over time. To stay a step ahead of fraudsters, you should use tools as advanced as theirs. 

But technology isn’t the only way to combat fraud. Human connection is also important, like you’ll find at the Global Insurance Fraud Summit. Here, you can collaborate and share information, including GenAI insights, with other insurance carriers and organizations, which is especially helpful to combat fraudsters targeting multiple carriers simultaneously.  

Remember: While GenAI is an impressive tool for detecting fraud, human oversight and connection are important components too. So, let’s fight fraud together. 

Authors

Learn more

Risk Detect: This digital platform helps identify and flag high risk activity with our digital risk detection and monitoring solution. Utilize advanced analytic algorithms and machine learning to get to key data points quickly and efficiently.