Uncovering the next wave of COVID-19 relief fraud through AI

Urupong/Getty Images

COMMENTARY | Many agencies are increasingly relying on AI to uncover fraudulent activities, and the emerging technology could hold the key to help recoup billions of dollars siphoned off from the federal government in pandemic relief scams.

During the peak of the COVID-19 pandemic, the US government allocated nearly a trillion dollars towards vital relief programs aimed at assisting American businesses and families. 

These programs offered forgivable loans to cover essential expenses. More than a dozen federal programs provided assistance in the form of direct financial payments for everything from housing subsidies and nutrition assistance to student aid for higher education. While millions benefited from these urgent relief efforts, the level of need and speedy distribution methods also opened the door to bad actors who took advantage of these programs to make fraudulent and illegitimate claims. From the beginning, many of these programs were streamlined to prioritize fund distribution, often with minimal oversight and controls, which left the programs susceptible to fraud. 

The scale of these fraudulent claims is only now becoming clear. As of April 2024, the Department of Justice has seized over $1.4 billion in stolen funds and charged over 3,500 defendants with crimes related to COVID-19 relief fraud, with associated losses of over $2 billion. Improper purchases included luxury real-estate, exotic cars, jewelry, lavish travel and cruises and designer clothing.  

There are likely billions more in COVID relief fraud to be discovered across agencies including the Small Business Administration, IRS, Social Security Administration, Department of Labor and others. To uncover more complex and undetected patterns, government agencies need to apply advanced technologies to identify those less apparent and subtle signs of fraudulent behavior. Many agencies are increasingly relying on AI to discern specific connections between individuals, businesses, addresses, and accounts for uncovering fraudulent activities and enhancing decision-making processes related to fraud detection and identification.  

AI enables public sector agencies to gain advantages over bad actors by using data more efficiently and effectively, ingesting massive amounts of data and building networks using broader contextual information. For example, the same person may be referred to using different names introduced from various data-entry protocols, errors, omissions, abbreviations, and even by intentional misrepresentation. These subtle changes make it difficult to identify a unique individual. By resolving references to specific entities including people, organizations, addresses, ID numbers, emails, and phones (known as entity resolution) and bringing in contextual data to understand the relationship between them, AI tools can help government agencies more accurately understand the complex relationships and links between entities. This helps to identify more advanced patterns of fraudulent behavior such as the use of shell companies, safe houses, and beneficial owners.  

Other governments around the world have already started to implement AI to identify complex cases of COVID relief fraud. In the UK, the Public Sector Fraud Authority operates as part of the government’s Cabinet Office and HM Treasury to understand and reduce the impact of fraud. In 2023, the PSFA began leveraging AI and decision intelligence to identify cases of fraud committed within the government’s Covid Bounce Back Loan Scheme. This approach has led to taxpayer savings of more than $300 million and equipped the UK government with full visibility into both the individual perpetrators of fraud as well as their entire network of connections to other enterprises and individuals to identify further fraudulent activity.  

For the U.S. Department of Treasury, the deployment of AI as a tool for check fraud detection has recovered over $375 million since first being implemented in 2022. Check fraud rates rose dramatically since the onset of the pandemic, due in part to counterfeiting, account takeovers, and physical theft of checks from the mail. An estimated 680,000 check fraud-related Suspicious Activity Reports were filed by banks and financial institutions where AI-fueled models have expedited fraud detection of Social Security payments, tax refunds and business checks with almost real-time visibility. 

With more pandemic-relief fraud charges continuing to emerge, AI holds the key to helping government agencies recoup billions of dollars in fraudulently claimed tax dollars. But it’s vital that these agencies go beyond current measures to adopt AI tools that can help uncover more complex patterns of fraud. Only by harnessing the power of AI to pinpoint intricate connections between individuals and businesses, uncovering fraudulent actions, and decisively stopping bad actors, can agencies truly confront the next level of fraud.