The Two Faces of AI

"Cost reduction, process optimization, improved efficiency, economic growth….” According to one of the world's leading insurance consulting companies, these are among the many great things artificial intelligence is bringing to the insurance marketplace.

This is probably correct, as the use of AI may be the most significant change ever for the world of insurance, and especially fraud investigation. Somewhere on the dark web, however, there is no doubt a similar webpage touting these same AI benefits for those seeking to commit insurance fraud crimes around the globe.

The fight against insurance fraud has always been about trying our best to keep a step ahead of those seeking to commit fraud. That race has never been easy, and while the dramatic rise of technology has brought incredible advancements to anti-fraud efforts, the “double-edged” sword of tech is also aiding fraudsters in devising new ways to commit their crimes.

In November, the sixth "Global Insurance Fraud Summit" in Singapore attracted an unprecedented number of attendees. Participants from international law enforcement, regulatory bodies, legal services, and the insurance industry frequently addressed the significant impact of AI on combating insurance fraud. According to Interpol, global financial fraud now surpasses $1 trillion annually and continues to increase. For perspective, only 19 countries worldwide have a GDP exceeding $1 trillion.

While AI is today’s buzzword, in reality it is not that new. In 1950, Alan Turing, considered to be the father of AI, published “Computer Machinery and Intelligence,” which proposed the Turing Test, a theoretical way to distinguish between a human and an AI. What has changed is the rise of the internet and expanding digital technologies, which have created the massive amounts of data essential for training AI models. Moore's Law states that computational power has doubled every 12–24 months for the past 50 years. Only in the last five years, though, has computational power increased to the level required to run today's deep learning algorithms, creating a new world of AI.

The most significant step forward is the rise of generative AI (gen AI). Gen AI is different from traditional AI, which is rule-based and has no way to learn from data or improve over time. In contrast, gen AI is able to learn from data and self-generate new data without human intervention. Gen AI is already in use to develop and write documents, engage in conversations, and even create avatars that appear to be fully human with the ability to speak, show facial expressions, and engage in conversations. While gen AI can learn news ways to identify fraud, it can equally learn new ways to successfully commit the same acts.

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. These tools use large datasets to train models to replicate a person's facial expressions, mannerisms, and vocal nuances with impressive accuracy. Such “deepfakes” create insurance fraud issues, such as a life insurance “policyholder” seeking to change a beneficiary or secure a policy loan, when in reality it is an AI-generated image or voice that is nearly indistinguishable from the real person. While these technologies are highly sophisticated, they still have imperfections detectable with the use of specialized tools.

When I began my insurance fraud investigation career, it was in a world of paper files, “high-tech” typewriters with backspace correction tape, 90-day status reports on claim files, and local SIU investigators working directly with area police and fire officials. Those days are gone. In today’s world, searchable data is the key to almost every insurance fraud investigation. The use of data is no longer the question, the issue is how to find and identify accurate and reliable data directly relevant to making the correct claim decision. Equally, we must understand that fraudsters today have access to the same data—or perhaps more since they are not concerned with laws, regulations, and privacy protections—to help them develop ways to commit new forms of insurance fraud.

The amount of data being created–and which AI feeds itself from–is staggering. Data scientists estimate 147 zettabytes (21 zeros after the number) of new data was created in 2024, but that is old news. In 2025, the estimate is data generation will increase to 181 zettabytes. International Data Corporation (IDC) released a report last year finding that approximately 90% of the world's data has been created in only the past two years.

According to the 2024 “Global Fraud Report,” 96% of fraud-prevention professionals are concerned  about the “industrialization” of fraud, and 79% have seen a significant increase in the sophistication of fraud attempts in the past year.” But how committed are insurance leaders, especially in the U.S., to truly fighting insurance fraud versus passing the cost along to consumers? This same study found “many fraud prevention professionals are still leaving the door open to fraudsters.”

To prove the point, the report found only 35% of respondents globally were investing in anti-fraud programs to stop fraud at the initial source, such as underwriting. And the U.S. appears to be among the worst at doing so with only 18% of respondents saying their organizations were spending resources to identify fraud at the start of the transaction versus later. This compares to rates for companies in Europe of 31% and in the Asia Pacific rim at 21%.

What is the biggest challenge we face in the fight against insurance fraud today? It is certainly not a lack of available anti-fraud technology. A visit to any insurance conference or trade show demonstrates that for sure. Remarkably the greatest challenge is no different than faced back in the world of paper files and in-person interviews. Money. How much are insurers willing to expend to protect themselves, their policyholders and consumers from the estimated $308.6 Billion lost to insurance fraud every year. If the amount of anti-fraud spending each year increased at even a fraction of the rate of projected data growth, we may well then find the real answer in how to successfully fight back against insurance fraud and crimes.

Matthew J. Smith is the executive director of the Global Insurance Fraud Summit, Inc. and president of Insurance Law Services, Inc. providing consulting and expert testimony on insurance matters.