Last month’s U.S. Government Accountability Office report confirms what many in government already know but have struggled to confront: improper payments are not an accounting anomaly—they are a systemic failure. With $186 billion in improper payments in FY2025 and cumulative losses exceeding $2.8 trillion, the federal government is hemorrhaging taxpayer dollars at a scale that should command urgent, structural reform.
This is not simply about waste. It is about a threat environment that has fundamentally changed.
Fraud today is faster, more scalable, and increasingly industrialized. AI-generated identities, synthetic documentation, and coordinated fraud rings are exploiting federal programs with a level of speed and precision that legacy systems were never designed to counter. From medical provider billing schemes impacting Medicaid to disaster relief fraud to unemployment insurance abuse, bad actors are leveraging AI while government systems remain anchored to a “pay first, chase later” model.
That model is no longer viable.
For decades, federal agencies have relied on retrospective controls—audits, investigations, and recovery efforts after funds have already gone out the door. But when losses reach into the hundreds of billions annually, recovery is no longer a strategy. It is damage control. Quite simply, the government cannot investigate its way out of this problem.
At the same time, agencies face a competing mandate: protect program access for legitimate recipients while safeguarding taxpayer dollars. Too often, outdated systems force a false choice between access and accountability—resulting in both massive fraud losses and frustrating delays for legitimate recipients.
In many public benefit programs, the biggest challenge is that caseworkers are overloaded, manual reviews are slow, and self-reporting alone often is not enough to keep records current. When systems are strained, payment error rates rise.
There is a better way, and it starts by shifting the point of decision.
Instead of chasing fraud after payment, the federal government must adopt a front-end risk screening and verification layer that assesses risk before money leaves the Treasury. Voice vetting tools are available now, alongside other AI-powered systems that can help agencies modernize payment integrity.
Viable AI-powered systems that can strengthen benefit integrity today:
- AI-driven risk assessment to screen claims and applicants: Machine-learning models can cross-reference applications in different states against known fraud indicators, identity inconsistencies, suspicious patterns, duplicate submissions.
- Voice analytics technology: Cutting-edge technology can quickly and accurately detect potential risk by analyzing responses to simple automated yes-or-no questions, helping agencies quickly clear most applicants while flagging a smaller subset for further scrutiny.
- Natural language processing (NLP) for ongoing monitoring: AI can scan claims narratives, provider submissions, case notes, complaint data, and inbound communications to identify emerging fraud trends, coordinated schemes, or suspicious language patterns.
The operational impact is straightforward but transformative.
Instead of treating every claim as suspicious—or worse, treating none of them as such—agencies can quickly clear the overwhelming majority of legitimate applicants while flagging a small subset for deeper review. That means more efficient processing for those who qualify, reduced backlogs for overwhelmed agencies, and a sharper investigative focus on the cases that may be problematic.
The opportunity is to improve integrity before payments go out—using technology to help agencies verify eligibility faster, reduce administrative burden, and move trusted applicants through the process more quickly. This challenge will only intensify as states prepare for new eligibility redetermination requirements for programs like Medicaid in 2027.
Front-end risk screening reduces improper payments before they occur, strengthens program integrity, and restores public confidence that taxpayer dollars are being protected—not simply tracked after the fact. Equally important, it aligns with a broader shift toward modernization and accountability in federal operations.
The GAO’s findings should be a forcing function. Continuing to rely on legacy controls in a rapidly evolving fraud landscape is not just inefficient—it is unsustainable. The scale of the problem demands solutions that operate at the same speed and sophistication as the threat.
The choice is clear: continue paying and chasing losses after the fact, or modernize how trust is established at the point of entry.
In an era of AI-enabled fraud, modern screening and verification technologies already being used in commercial environments provide a credible, proven way to protect taxpayer dollars while ensuring that those who depend on federal programs receive support without unnecessary delay.
The status quo is costing too much. It’s time to fix it before the next trillion is lost.
Donald Blersch is a former senior public-sector executive with more than 35 years of experience and current Chair of the INSA Security Policy Reform Committee. He wrote this for InsideSources.com.
Reprinted with permission from DC Journal by Donald Blersch.
The opinions expressed by columnists are their own and do not necessarily represent the views of AMAC or AMAC Action.