Washington CORE recently supported Nikkei Digital Governance in publishing an article examining how the U.S. government is using artificial intelligence to detect and prevent improper subsidy payments. The feature explores how federal agencies are applying AI and advanced data analytics to identify administrative errors, fraud, and duplicate payments across large public benefit programs.

Improper payments in the United States have reached significant levels in recent years. According to U.S. Government Accountability Office, improper spending peaked at approximately $281 billion during the COVID-19 pandemic and still totals roughly $162 billion annually. To address this challenge, the U.S. Treasury has expanded tools such as the “Do Not Pay” portal, which uses AI and cross-agency databases to verify eligibility and flag suspicious transactions before funds are distributed. The system integrates data sources such as death records, criminal databases, and tax information to prevent payments to ineligible recipients.

The article also highlights the growing role of AI platforms, including technologies developed by companies like Palantir, in supporting large-scale government data analysis while raising debates around privacy and oversight.

The full article is available by paid subscription through Nikkei Digital Governance. You can read more about it here in Japanese.