Curbing Federal Fraud: Shifting from Pay-and-Chase to AI-Driven Prevention

Approximately $233 to $521 billion in federal funding is lost to fraud every year according to a 2024 Government Accountability Office (GAO) analysis. To curb this alarming trend, a significant shift in federal strategy is being advanced: moving away from the pay-and-chase model toward a proactive, pre-payment detection approach powered by artificial intelligence (AI) and advanced data analytics. The shift was discussed in a January hearing of the House Subcommittee on Government Operations titled “Curbing Federal Fraud: Examining Innovative Tools to Detect and Prevent Fraud in Federal Programs.” 

The Move Towards Proactive Detection

Citing recent instances in Minnesota where the lack of guardrails allowed bad actors to enrich themselves at the taxpayers' expense, Chairman Pete Sessions, a Republican Representative for Texas, emphasized that "fraud should be detected before it happens."  To combat the pervasive fraudulent use of federal funds, participants of the hearing championed human-supervised AI tools designed to find fraud before funds go out the door.

Members of several key federal entities testified as witnesses, including:

• Ken Dieffenbach, Executive Director of the Pandemic Response Accountability Committee (PRAC), which was established by the CARES Act to provide oversight of pandemic assistance funds but has since been expanded to other agencies.

• Renata Miskell, Deputy Assistant Secretary for Accounting Policy and Transparency at the U.S. Department of Treasury’s Bureau of Fiscal Service, which has the mission of promoting financial integrity and operational efficiency of the U.S. government through exceptional accounting, financing, collections, payments, and shared services.

• Dr. Sterling Thomas, Chief Scientist at the U.S. Government Accountability Office, which acts as an independent nonpartisan congressional watchdog.

Key federal entities presented their initiatives:

• The PRAC: Now extended through 2034, the PRAC is developing an AI-enabled fraud prevention engine. This tool is trained on 5 million pandemic applications and can analyze 20,000 applications per second to identify anomalies and hidden connections before payments are issued.

• The Department of the Treasury: The Treasury is expanding and enhancing the Do Not Pay (DNP) system, a government-wide tool to detect and prevent improper payments. While only 4% of Federal programs could access all available DNP data in Fiscal Year 2024, all Federal programs are on track to fully use the system by the end of Fiscal Year 2026. In 2024 the Treasury developed a team that identified new data sets and machine learning, and AI techniques that could have prevented an estimated $28 billion in improper payments had it been used the previous year. The Treasury is also working to improve its payment verification screening processes and is requesting (in a privacy-preserving manner) expanded access to taxpayer identification, income and other data to do so.

• The GAO: The GAO highlighted its Fraud Risk Framework, recommending that agencies design risk management directly into their programs and build analytics-based fraud risk analysis teams. GAO Chief Scientist Dr. Sterling Thomas, while optimistic about AI innovation, cautioned that “rapid deployment without thoughtful design,” has already led to unintended outcomes. He stressed the importance of data quality, warning that AI outputs are only as reliable as the data used to train and inform them. Referencing the GAO’s AI accountability framework, Dr. Thomas advised employing solutions that achieve desired outcomes with the least amount of complexity. He indicated that while AI is a powerful tool and has great promise in preventing fraud, the enhanced use of traditional data analytics tools and other methods also has the potential to drastically reduce fraud.

Overcoming Barriers and Modernizing the Workforce

Despite these innovative advancements, witnesses identified several obstacles, including limited access to certain datasets for the DNP system and a lack of a workforce trained in employing AI and data analytics. To address this, Dr. Thomas suggested creating a Digital Services Academy to ensure the federal workforce has the necessary skills to employ AI effectively and in a manner that protects data privacy. He also suggested that Congress establish a permanent analytics center of excellence to encourage innovation and distribute fraud prevention tools.  

While the hearing underscored the necessity of these tools for protecting billions in taxpayer dollars, concerns were raised regarding data privacy and the potential for these massive databases to be misused.

Aligning with Federal Priorities: Vander Weele Group Solutions

At the Vander Weele Group, we believe that AI is a powerful tool to identify risks, but that technology is inadequate on its own. AI results need to always be human-vetted and validated, not only to establish accuracy and align with actual field conditions but to withstand legal scrutiny.

The Vander Weele Group offers established methodologies, advanced technology tools and deep grants expertise that align with this new approach, backed by decades of experience in the fraud control arena.

• Agent Tipster™: Reflecting the administration's focus on AI, Agent Tipster™ is an AI-powered risk assessment tool designed to analyze large data sets. It utilizes machine learning models to ingest and analyze vast troves of data, pinpointing high-risk anomalies and irregular trends. Our multi-disciplinary teams of auditors, fraud experts, grants specialists, and attorneys then validate the findings and, in extreme circumstances of criminal misconduct, prepare prosecution-ready reports that can withstand scrutiny in court.

• Guardrails 360™: This comprehensive platform provides a 360-degree perspective on grant oversight. It sets up critical safety checkpoints across more than 700 grant lifecycle activities, preventing compliance and operational issues before they occur. By mapping legal and regulatory requirements and correlating each standard to the appropriate activity, Guardrails 360™ ensures that programs are compliant and secure. Our team provides an added measure of safety by inspecting fraud prevention measures such as internal controls, applying expert analysis to data, and asking thoughtful questions across the stakeholder ecosystem. We bring creative policy solutions to prevent, detect, and resolve matters involving fraud and abuse.

• DTS Navigator™: As federal oversight tightens under Executive Order 14222, the Defend the Spend (DTS) verification mandate is changing the landscape of state draw-down requests on federal grant funds. Even as challenges play out in the courts, the risk is clear: insufficient data can lead to frozen funds, cash flow delays, or even cancellation of grants. DTS Navigator provides the technical and human infrastructure necessary to help you navigate complex Payment Management System (PMS) workflows. We support your grant teams in building justifications for every drawdown to prevent cash flow delays. Our data team ensures your proof of service meets the latest federal requirements, and we handle communication and technical assistance directly with your subrecipients, checking for compliance at every level.

Watch our webinar, "Detecting Fraud in Grant Programs Through AI and Data Analytics," to learn how to:

• Synergize Tools: Understand how traditional analytics and AI work best together to uncover complex schemes.

• Identify Trends: Identify practical algorithms and analytic techniques to spot irregular patterns and anomalies in massive datasets.

• Find Fraud: Learn how detailed data and document analysis can expose patterns of fraud, waste, and abuse in your grant datasets.

---

To learn more about Vander Weele Group and our full suite of grant oversight services, visit VanderWeeleGroup.com or contact us at 773-929-3030 to consult with a grants oversight expert.