AI has been used to deceive people in everything from phone calls to voters to faking celebrity gifts. Now, the US Treasury is claiming that machine learning AI has played a key role in enhancing its fraud detection processes over the past year. If a broken clock can be accurate twice a day, an AI might do something good once.
In a new announcement, the Treasury Department said it prevented and recovered more than $4 billion worth of “fraud and improper payments” in the last financial year (October 2023 to September 2024). This figure represents a significant increase from the previous year, when it amounted to just $652.7 million. A quarter of the $4 billion is expected to come from recoveries from “machine learning AI to accelerate identification of Treasury check fraud.” Again, is it like making a deal with the devil? yes. But that’s 2024.
The $1 billion includes $2.5 billion for prevention by “identifying and prioritizing high-risk transactions” and another $680 million for additional prevention technologies.
The Treasury Department plans to share the technology with other federal agencies, but some have already implemented their own technology. For example, the IRS is taking steps to use AI to uncover tax evaders, automate services, and conduct audits.
