The US Treasury Department’s bet on artificial intelligence (AI) to fight against financial crime is paying off.
Machine learning AI helped the US Treasury search a massive volume of data and helped recover $1 billion worth of cheque fraud in fiscal 2024, nearly three times what the Treasury made in the previous fiscal year.
The treasury department credited AI, which deducted and recovered more than $4 billion in unauthorized transfers worth of fraud in the fiscal year 2024.
Renata Miskell, a top Treasury official, stated that the use of artificial intelligence has been transformative for the agency, but it required a human to label it as fraud.
Miskell alleged that fraudsters nowadays are becoming adept at hiding and trying to secretly manipulate the system.
He credited AI, which helped officials prevent and recover by finding hidden trends and abnormalities and avoiding fraud detection and prevention.
Miskell emphasized that even though artificial intelligence can identify questionable transactions, humans can ultimately decide whether a transaction qualifies as fraud.
Artificial intelligence can be tremendously helpful in fighting against financial crime by searching through endless data and figuring out subtle patterns in a fraction of time, which would take humans many hours to do the same. According to experts, advanced artificial intelligence can find suspicious transactions in milliseconds.
It is particularly crucial for the Treasury, one of the biggest payers in the world.
US officials are using AI covertly to detect financial fraud in late 2022, taking inspiration from many banks and credit card companies that currently use AI to stop criminal activity.
Miskell declined to provide specifics to prevent disclosing the identities of corrupt individuals. He stated that the Treasury is investigating ways to modify its fraud-detection techniques using the methods used by major banks and credit card companies.
The objective is to safeguard public funds from fraudulent activities, which became more prevalent during the COVID-19 pandemic when the government was rushing to provide emergency assistance to individuals and companies.
Treasury is not using generative AI, the one that captures the users of Google‘s Gemini and OpenAI’s ChatGPT by creating visuals, writing song lyrics, and providing answers to complex questions.
Fraud detection efforts focus on machine learning, a kind of artificial intelligence that is particularly good at analyzing large volumes of data and making predictions and judgments based on trained knowledge.
The Treasury provides 1.4 billion payments worth around $7 trillion to 100 million people annually. It sends out stimulus checks, tax refunds, Social Security and Medicaid benefits, and provides paychecks to federal employees.
The crucial position of the Treasure makes it a prime target for scammers looking to steal from taxpayers.
The Internal Revenue Service said it has used artificial intelligence (AI) to go through complicated tax filings from hedge funds, legal companies, and other sources to find tax evaders.
Juniper Research predicted that online payment fraud would exceed $362 billion by 2028, and artificial intelligence would contribute to some of those frauds.
The police from Hong Kong reported a notorious example from earlier this year. A finance employee had to pay $25 million to criminals who were tricked through a deepfake video.
US officials are worried that artificial intelligence will bring new risks to the financial system. In June, Treasury Secretary Janet Yellen alerted bankers that artificial intelligence in banking would pose serious risks.
Last year, top regulators, led by Yellen, classified artificial intelligence as a new danger to the financial system.
According to a Treasury spokeswoman, the government is working more quickly to improve the fraud-detection resources available to state and federally-sponsored programs. Officials are experimenting with new data sources and collaborating with state organizations to fight unemployment insurance fraud and detect fraudulent activity and questionable payments.