A.I. helped the feds catch $1 billion of fraud in one year. And it's just getting started

The U.S. Treasury Department has significantly enhanced its efforts to combat fraudulent payments by leveraging artificial intelligence (A.I.) to identify and prevent suspicious transactions. The introduction of these A.I. tools has already proven transformative, leading to over $4 billion in fraudulent payments in fiscal year 2024, a six-fold increase from the previous year's recovery of approximately $652 million.

The Treasury implemented its AI-based fraud detection system at the start of the 2023 fiscal year. This system alone was responsible for recovering $1 billion in check fraud during 2024—nearly triple what was recovered the prior fiscal year. Treasury officials attribute this dramatic improvement to A.I.'s ability to efficiently analyze large amounts of payment data and identify high-risk transactions in milliseconds. This task would be impractical for human auditors to handle at the same speed.

"A.I. has been transformative," said Renata Miskell, a senior Treasury official. "Leveraging data has enhanced our ability to detect and prevent fraud."

This sophisticated A.I. model uses machine learning techniques rather than generative models like OpenAI's ChatGPT or Google's Gemini. Its focus is on pattern recognition, anomaly detection, and predictive decision-making. The Treasury Department's efforts reflect similar fraud detection practices employed by leading financial institutions such as banks and credit card companies.

The deployment of A.I. tools resulted from a spike in fraudulent activities that coincided with the COVID-19 pandemic when the federal government disbursed emergency funds. This massive outflow of aid presented new opportunities for fraudsters, accelerating the need for AI-based defenses.

The Treasury processes over 1.4 billion payments annually, valued at approximately $7 trillion—including Social Security benefits, federal payroll, tax refunds, and stimulus checks. This makes it a key target for cybercriminals seeking to exploit vulnerabilities in the payment system.

Notably, the Treasury's A.I. tools have prevented fraud and streamlined payment scheduling processes, recovering an additional $180 million.

While A.I. plays a critical role in spotting suspicious transactions, human oversight remains a crucial part of the process. Treasury officials confirmed that flagged transactions are subject to review by federal staff before final determinations are made. This balances A.I. efficiency and human judgment, reducing false positives and preventing misuse.

The Treasury has also collaborated with law enforcement agencies, leading to several arrests and ongoing investigations. The department is actively working with state agencies to combat unemployment insurance fraud and enhance the fraud-detection tools available to federal and state programs.

Though AI helps detect fraud, it also presents new challenges. Treasury Secretary Janet Yellen has cautioned financial institutions about the potential risks of AI in the financial sector, calling it an "emerging vulnerability." AI has already been implicated in various fraudulent schemes, such as deepfake technology, which was used to trick a finance worker in Hong Kong into transferring $25 million to criminals.

TaxGPT: Advancing AI And The Tax Profession (forbes.com)

U.S. regulators have acknowledged these risks and are working to develop safeguards to protect the financial system from abuse. However, the Treasury's ongoing A.I. efforts signal a proactive stance, with plans to expand further and refine its fraud detection methods. Officials remain cautious about revealing too many details, fearing it might "tip off bad actors."

As the Treasury accelerates its adoption of AI-powered tools, it aims to stay ahead of rapidly evolving fraud schemes, significantly as online payment fraud is projected to exceed $362 billion by 2028. The department continues testing new data sources and enhancing its detection methods to remain one step ahead of fraudsters.

The Treasury's early success demonstrates how A.I. can protect taxpayer money and improve operational efficiency when adequately integrated. With AI technology still in its infancy within the federal government, the Treasury sees significant potential for further improvements as it refines its approach and learns from financial institutions leading the way.