“Curbing Federal Fraud: Examining Innovative Tools to Detect and Prevent Fraud in Federal Programs”
No chapters available yet
Legend
Transcript markers
- text Sentence being played
- text Search result
- text Intervention cited by the link you opened
Click a sentence to play the video from that moment. Select text to quote it with the speaker, timestamp and link.
More options
House Oversight and Government Reform Subcommittee on Government Operations U.S. House of Representatives
Follow its sittingsSpeaker 1
So I will say to him today, thank you for being part of this effort. Thank you for the opportunity for us to approach getting to difficult subjects with the same expertise, the same knowledge, and close to the same answer. I will yield my time and now yield to the distinguished gentleman, the ranking member. Gentlemen, recognized.
Speaker 2
Thank you very much, chairman Sessions. I don't know that you're going to cost me any votes. There used to be an old practice around here by some of the old timers who'd walk up to you before an election and say, I can be for you or against you. Which woman do you want? So you continue to be for me.
I appreciate that. I really do. And I wanna thank you for working together to hold this hearing today. Obviously, I wanna thank the witnesses that are here before us, the members of this committee who have joined and will be joining. Although it's captured the news cycle with all sorts of attention in recent weeks, the problem of fraud and improper payments in the federal government programs has existed for many years as we all know.
It has bedeviled presidential administrations on both sides of the aisle, and it has cost taxpayers hundreds of billion dollars. In fact, a 2024 government accountability office analysis estimates that the federal government loses a staggering $233 to $521,000,000,000 in fraud every year. That's just absolutely amazing, and it's beyond comprehension. So Chairman Sessions and I have worked on a bipartisan basis, as you've heard, to address this issue, and we've done it for the past few years. We remain keenly focused on the task because every dollar lost to fraud, we all know, is prevented from going into those desperately needed programs of assistance that the federal government and its programs provide.
The government's assistance programs play, I believe, an irreplaceable role in helping small businesses to grow, keeping children fed, ensuring access to affordable health care, and protecting so many other sectors of our society. The fight to combat fraud must not malign these programs and rip away support from people that need it, but instead, we've gotta find and make sure that our laws are faithfully executed, and that taxpayer dollars go to the people they were meant to serve. As our committee discusses the fight to detect and prevent fraud before it happens, it's extremely important to reflect on the work that has been done and much of the work that hasn't worked, if I can use that terminology. At the beginning of the year, the so-called department of government efficiency sought to go it alone, ignoring prior recommendations for cost savings, for fraud prevention, and for efficiency from our federal oversight bodies. Firing scores of federal employees and unilaterally destroying government programs that, Mr. Musk did not personally like, really led to much of the damage, I believe, that we can look back on today and pinpoint.
When Doge employees failed to identify actual instances of rampant race, fraud, and abuse, they made up, in their own way, their own numbers and logged wildly inaccurate claims of their so-called, quote, wall of receipts that we've all read about. Now here we are a year later, and government spending has actually increased. Let me repeat that. One year after, that government spending has actually increased. And so the fact is that our government already had in place fraud-fighting expertise instead of illegally defunding the council for inspector generals on integrity and efficiency, and firing federal workers trained in fraud prevention.
The focus went on funding these other offices and doing away with career civil servants and inspector generals that truly understand the systems that we hold them accountable for. So as we move beyond the chaos and move beyond some of the destruction of the last year, we've got a look at the tools of the future that prevent fraud before it happens. I am greatly encouraged by the progress of the pandemic response accountability committee in creating analytic systems to prevent and detect fraud. Progress, by the way, has already recovered over $500,000,000 in taxpayer funds. Prac's model, as the chairman has said, demonstrates the belief and the reality that the combination of data sharing between siloed federal agencies and responsibly implemented human-supervised artificial intelligence systems can stop fraud before it happens.
AI systems working with large databases and datasets can detect patterns and connections with fraudulent actors that humans often can't flag. Certain applications that can't be processed correctly are paused and reviewed by a human before payments happen. Of course, as we all know, and I think as we will all agree, the use of any artificial intelligence systems with private data of Americans requires absolute caution, and fraud fighting officials must receive proper training to ensure that these models are not trained on high quality data that could be used in the wrong way. An unreliable AI system or an AI system without a skilled workforce to train and use it would be worse than no system at all. So before I conclude, I just wanna address the harmful politicization of fraud-fighting efforts that we've seen taking place.
Legitimate efforts to combat fraud in federally funded programs must never result in cuts to programs that Americans rely on. We can do better than that. Just last week, the administration froze $10,000,000,000 in social services for the states of New York, California, Colorado, Illinois, and Minnesota without any real clear reasoning, impacting hundreds of families and hurting the poorest among us, but not as I've been able to detect preventing any sort of fraud. So politically motivated attacks accompanied by announcements like that the assistant attorney general for fraud prevention made the other day, apparently, lead to the kind of unfortunate reporting that we have. I think that we've gotta be able to do what we do in a real clear and concise way, that we've gotta be able to share that information, that this cannot be a partisan fight.
The people who are hurting as a result of fraud, waste, and abuse come from all parts of this country. They're all backgrounds and religions. They can help themselves and others cannot, but we have the power in this committee and indeed this Congress to put in place the sort of things that work. So I wanna commend the work of Prac. I wanna thank the chairman for being and having a long vision on this and working together in a bipartisan way so that members of this body and members of the house of representatives recognize that there is a good, clear, and workable way out of this if we can put aside those things that separate us long enough to be able to deal with the fact that 233,000,000 to 531,000,000,000 are going out the door every day while we engage in some of those battles.
So I want to yield back, mister chairman. I wanna thank you again for this. I think it's fair to say that everybody on the committee has a clear desire to really get their arms around this. It's been a lonely couple of years working on this, quite frankly. But I think people now are starting to recognize that once we do something and it is verifiable and we do it again, we can make a real and lasting difference.
So I thank you, and I yield back any time I might have.
Speaker 1
Jimin yields back his time, and I wanna concur with him that perhaps the greatest thing we can do in this is to get ideas from the people who are closest to it, who studied it, who understand it, who understand the characteristics and the relationships inside government of maybe not just the limitations, but the frailties that keep us from getting closer. I wanna thank you and the gentleman for his opening statement. Today, I'd like to note that we have the young chairman of the entire committee, Mister Comer, who's with us today. Normally, he's on this hot seat over here. But, Jim, I want to welcome you to this subcommittee, and thank you very much.
I'd also, without objection, like to waive on congressman Timmons of South Carolina and Wickenshaw of Virginia who are waived on the subcommittee for the purpose of questioning the witnesses at today's subcommittee. It's obvious I'm hearing discussion from the gentleman from Tennessee who wanted me to mention his name also. So I'm delighted that each of the members are here today. We're gonna move directly to our witnesses. And so today, I'm pleased to introduce the witnesses who really today will be the stars of the show.
And I've instructed them, as you know I always do, to please come and tell us not just the information that we need, but the story behind it. Because I think the compelling evidence that they bring needs some clarification about really what a difference it makes. And so today, I'm pleased to introduce Ken Diefenbach. He is executive director of the pandemic response accountability committee known as the PRAC, and he really gets the gold star. He is a star witness, he has been with us before.
And as I told him last time, he did well enough to get invited back, and he's taken us up on today. So he leads the committee's efforts to support and coordinate oversight of the trillions of dollars that were spent on pandemic response and help to understand the detection of fraud, waste, abuse, and management of related funds. Ken, thank you for being here today. Second, we have Renal Miskow. She's the director assistant secretary for accounting policy and financial transparency at the US Department of the Treasury.
I think this is a new role for her, and we're delighted. I think that because of the great work that she's done, she is also a valuable asset to this interest. There, she leads work to safeguard taxpayer dollars by modernizing payments, preventing fraud, reducing improper payments, and promoting fiscal responsibility. And she told me yesterday that this goes all the way to the top, to the secretary, that the secretary is interested and sees this as being part of President Trump's management agenda. Last, Sterling Thomas is the chief scientist at the GAO, the Government Accountability Office, where he leads the agency's science and technology work and assesses and evaluates emerging technologies for application in government, including AI, artificial intelligence.
And I enjoyed being with him yesterday. I had to leave a little bit early, but I really appreciate it. Doctor, thank you so much for taking the time to join us. So pursuant to committee rule nine g, I will ask that the witnesses rise and raise their right hand. To be sworn, do you solemnly swear or affirm that the testimony you're about to give is the truth, the whole truth, and nothing but the truth?
So help you, God. Thank you. Let the record reflect these witnesses answered in the affirmative. Thank you. May take your seat.
The opportunity for us to have you give testimony today to the United States Congress. This subcommittee is important to us. We will count on that you have. Let me remind the witnesses that we have read your written statement. We talked about it yesterday, and it will appear in the hearing record in full.
So while we do talk about the five minutes, I instructed each of you yesterday to tell your story, and to remember that I have a slow gavel, and mister Infume agrees with that. We want to hear from you. This is important. But I would ask that you also pay attention to the green light, red light, yellow light, and red light. And I would now go to Mr. Dieffenbach for his opening statement.
Gentlemen's recognized.
Speaker 3
Thank you. Chair Comer, chair Sessions, ranking member in Fume, members of the subcommittee, it is an honor to be here today to discuss the Pandemic Response Accountability Committee, or Prac, and our work to investigate fraud and improve fraud prevention across federal programs. Thanks to Congress and to this subcommittee, the One Big Beautiful Bill Act extended the prac and its data analytics capabilities until 2034, provided $88,000,000 in funding, and expanded our jurisdiction to programs funded in the law. The prac team is thankful for this opportunity to demonstrate the value and effectiveness of our fraud prevention work. Operating on an annual budget of 18 and a half million dollars a year, to date, the Prac has helped recover over $500,000,000 for the taxpayer.
As executive director of the Prac, an entity Congress created to oversee over $5,000,000,000,000 in relief funding, I work with a phenomenal team that is leveraging artificial intelligence to collect, organize, and analyze data to rapidly provide insights into fraud risks. This proactive approach is clearly needed so that fraud is prevented before funds are disbursed. This past summer, the prac issued two fraud alerts that identified over 79,000,000,000 in potential fraud that could have been prevented with pre-award vetting and cross-agency collaboration. To address these issues, the prac is developing an artificial intelligence-enabled fraud prevention engine. Trained on 5,000,000 pandemic applications and other data, the tool can quickly identify anomalies, trends, patterns, and hidden connections in future applications before payments are made.
Had our fraud prevention engine been in existence in March 2020, pre-award vetting would have flagged at least tens of billions of dollars in fraudulent claims for further scrutiny, allowing agencies to prevent fraud. The prac is also actively engaging with partner inspectors general to identify opportunities to prevent fraud in programs funded in the One Big Beautiful Bill Act, and to address fraud risks beyond identity theft and eligibility fraud issues or issues such as traditional procurement fraud. Of particular focus will be cross-program risks, as fraudsters rarely target just one government program. They exploit vulnerabilities wherever they exist. The prac also works with IG offices to develop analytics tools that provide new insights and improve their effectiveness, in their oversight mission.
For example, we developed a risk dashboard for the Pension Benefit Guaranty Corporation, OIG, that to date has contributed to their recovery of over $260,000,000. And we just launched a dashboard for the Federal Communications Commission, OIG, that's focused on four programs that dispersed $11,000,000,000 to 14,000 different entities. The Prac also provides investigative support to more than 50 law enforcement partners who related to over 1,200 investigations, with over 24,000 subjects with potential fraud losses of over 2 and a half billion dollars. We also partner closely with GAO and the treasury department, and it's important to note that the Prac and the treasuries do not pay program are complementary platforms that both work to protect taxpayer dollars from different angles. The PRAC focuses on a broad set of risks, patterns, trends, anomalies, and hidden connections among data from a wide array of programs, applications, and transactions.
Our access to law enforcement sensitive and other unique data, our 119,000,000 pandemic aid applications, and over 127,000 known pandemic fraud cases or suspected cases can reveal powerful new insights and serve as an early warning system of organized, often transnational, criminal conspiracies and other emerging threats. In one of the tens of thousands of pandemic fraud cases, a Prac investigation identified one scheme involving more than 450 applications from over 100 different applicants across 24 states. This is but one example where the proactive use of data and technology could have prevented or aided in the early detection of a scheme, mitigated the need for a resource-intensive investigation and prosecution, and helped ensure taxpayer dollars went to the intended recipients and not the fraudsters. As many of you already pointed out, every dollar that goes to a fraudster does not go to the recipients the congress intended to help, the small businesses, the unemployed individuals, veterans, just to name a few. With the support of Congress and the PRAC's talented staff, we will continue our work on behalf of the taxpayers to investigate fraud and demonstrate the value and effectiveness of fraud prevention.
Thank you again for your continued strong support of the prac, the IG community, and independent oversight. This concludes my prepared remarks, and I look forward to your questions. Thank you.
Speaker 1
Thank you very much. Yesterday, you spoke about the pride of authorship that you have for your job and how other people who may be associated with it saw that. I hope they're having a chance to at least get a tape of you today.
Speaker 1
Perhaps they are. Well, the gentleman is now recognized for an opening statement.
Speaker 5
Thank you, Chair Comer, Chair Sessions, Ranking Member Mfume, members of the subcommittee. Thank you for the opportunity to share how Treasury is supporting federal programs in preventing fraud and improper payments. Treasury is firmly committed to safeguarding taxpayer dollars and advancing data-driven solutions to prevent fraud and improper payments before they occur. Each year, the treasury, on behalf of federal agencies, disperses trillions of dollars in payments. Ensuring those payments go to the right people in the right amounts and at the right time is both a matter of fiscal responsibility and a matter of public trust.
Under Secretary Besson's leadership, Treasury is modernizing stewardship of taxpayer dollars. At the core of the treasury's efforts are two initiatives. First, expanding the use and utility of DoNotPay, and second, enhancing Treasury's payment verification processes to flag risky payments. I'll provide an overview of both initiatives, followed by the challenges the treasury faces to implement them. Do not pay is a government-wide tool provided by Treasury for agencies and states operating federal programs to detect and prevent the leading causes of improper payments.
As an analogy, think of the treasury as America's bank. Across the country, certifying officers, grant managers, caseworkers are like bank tellers, responsible for ensuring payments go to eligible individuals and entities. Do Not Pay is a tool that helps these frontline workers detect risk when making awards and certifying payments. Despite its promise, DoNotPay has faced two challenges. First, the program has been underutilized.
In fiscal year 24, only 4% of federal programs could access all available data. Second, Do Not Pay has not had sufficient authority to access key federal databases that could detect the most common drivers of improper payments, namely verifying identity, financial status, and death. To address these challenges, in January 24, the treasury created a tiger team to study the issue and prototyped innovative solutions. The Tiger team identified promising new datasets and machine learning and AI techniques that could have prevented about $28,000,000,000 in improper payments if it was used in the prior year. In March 2025, the president issued executive order 14249 protecting America's bank account against fraud, waste, abuse.
The executive order embedded many of the lessons learned from the TIGER team. And since then, Treasury, in partnership with the Office of Management and Budget, has made substantial progress in expanding and improving Do Not Pay. By the end of this fiscal year, all federal programs are on track to fully utilize Do Not Pay. We're also working to add new high-value datasets and are overhauling our technology to deliver more useful results. In addition, the treasury is enhancing its payment verification screening to identify anomalies prior to agency certification.
The screening helps ensure that payments have valid accounting codes, that the bank account provided is open and belongs to the payee, and that payments are not going to deceased individuals. While this progress is significant, additional statutory authority would help the treasury fully achieve its objectives. I'm grateful for congress's support in granting treasury permanent access to the full death master file through passage of S 269 last evening. Treasury is also seeking limited access for Do Not Pay to validate taxpayer identification numbers and income in a privacy-preserving manner that is consistent with Do Not Pay's rigorous access controls. These data sources would dramatically improve eligibility determination and fraud prevention.
In closing, I look forward to working with this committee to help ensure federal dollars reach the people and entities that they are meant to serve. Thank you, and I look forward to your questions.
Speaker 6
Chair Comer, chairman Sessions, ranking member Infame, and members of the committee, thank you for inviting me to participate in today's hearing to discuss how innovative tools can be used to detect fraud in federal programs. As GAO's chief scientist and throughout my career in industry, academia, and the intelligence community, I have seen great advancement in data science. These innovations offer exciting opportunities to improve government efficiency. As you know, GAO, we're a nonpartisan watchdog for Congress. We have expanded our science technology team in recent years. It includes a group of data scientists.
We know firsthand from their work that AI holds great promise in furthering GAO's mission and your goals of safeguarding the taxpayers' dollars. My aim today is to offer three important actions that will help us reach that goal. First, we must continue and augment our traditional anti-fraud efforts. Second, we must apply AI thoughtfully and ensure that we use quality data to mitigate its well-known and well-described risks. And third, we must ensure the federal workforce has the skills they need to apply new innovations like AI.
Regarding my first point, GAO is a large body of work on fraud and improper payments in the federal government. We found the federal government reported an estimated $162,000,000,000 in payment errors or improper payments during fiscal 2024. And that's almost certainly an underestimate, because it doesn't include estimates for some federal programs. We've outlined numerous ways that Congress and federal agencies can tackle this problem with existing capabilities, such as reducing data silos. For example, Congress could make permanent the Social Security Administration's authority to share its death list with Treasury's Do Not Pay system.
According to treasury, just one year of access to this data resulted in total net benefits of $109,000,000. But treasury's access to this data was set to expire in December, although I learned that it just passed for permanent expansion, which is great news to us. These traditional methods routinely prevent and detect fraud. By enhancing them, we can save taxpayer dollars today without new technology. On my second point, like I said, I'm optimistic about innovation using AI, but we must be thoughtful about it.
In data science, we often say garbage in, garbage out. Nowhere is that more true than with AI and machine learning. If we start trying to identify fraud and improper payments with flawed data, we're going to get poor results. AI is still in its early stages of development and implementation, and rapid deployment without thoughtful design has already led to unintended outcomes. Before pouring data science on the problem, we need solid, reliable, ground truth data and a human in the loop to ensure that data reliability and the application of the technology.
GAO has an AI accountability framework, which lays out these and other principles. One piece of advice that emerges from such principles is to find a solution that produces the desired result with the least complexity. For example, in response to our recommendations, the Small Business Administration screened all the PPP or Paycheck Protection Program loans made before December 2020 with a rules-based tool just looking for indicators and not using AI. And they still identified $4,700,000,000 in loans that went to ineligible recipients or were used for unauthorized purposes. To build on that foundation, we need more innovation in government.
We recommend one way to do this is that congress could establish a permanent analytics center of excellence like the Prac we've been talking about to distribute tools to the community that more efficiently and effectively identify and prevent fraud and improper payments across the government. My final point is that harnessing innovation also requires a federal workforce that has the right skills. But agencies continue to face barriers in hiring, managing, and retaining staff with these advanced technical and data-science skills. This is another area where GAO has made recommendations, and we have explored innovative big picture ideas like establishing a digital services academy. Rapid advances in AI capabilities hold great promise to reduce fraud in federal programs.
We at GAO believe that the federal government can and must realize those opportunities and do so upon a firm foundation of reliable data and a digitally skilled workforce. Chairman Comer, chairman Sessions, ranking member Infame, and members of the committee, thank you, and I'd be happy to respond to your questions.
Speaker 1
Doctor Thomas, perfect, five minutes. That's a 10.0. So, congratulations. I'd now like to go to questions. We would go to the distinguished gentleman from Kentucky.
The gentleman is now recognized, Mr. Chairman.
Speaker 4
Thank you, Chairman Sessions, and record member Mfume for always delivering on good quality subcommittee hearings. Last week, the full committee held a hearing to examine the massive fraud in the Minnesota social services program that resulted in at least $9,000,000,000 in taxpayer funds being wasted. Much of this fraud occurred because there were no proper safeguards put in place to prevent fraud before the funds were paid. There was also too little oversight on how the money was used once it went out the door to the recipients. During the Biden administration, Democrats rushed out federal funds with virtually no safeguards, in a massive theft of taxpayer dollars.
Republicans repeatedly warned that the absence of guardrails would invite waste, fraud, and abuse. Identifying fraud before money goes out the door is necessary to prevent any further repeats of these failures and to protect taxpayer dollars. We need to get ahead of the criminals, continue to help the Department of Justice in arresting, prosecuting, and jailing those responsible, and ensure federal programs serve those who are truly in need. This committee, hopefully, is serious about this. I know, the majority is on the committee, and we plan on continuing to identify waste, fraud, and abuse and hold people accountable.
My first question for, mister, Dysonbach. How could practice tools and analysis have been used to stop the large identity and eligibility schemes that defrauded programs in Minnesota?
Speaker 3
Thank you, Chair Comer. And, again, thank you for your support. The hallmark of most fraud schemes is that people hide information. So leveraging data analytics allows us to see patterns, trends, anomalies, hidden connections, to shine a bright light on what's actually happening. That is the path forward.
So we have to assemble the right data, the right team, the right tools, which we already have at the Prac. Thanks to your support. We just need to think more about the jurisdiction of how we're employing those tools, but data is the solution. Absolutely.
Speaker 4
K. Miss Miskill, given the treasury is the last stop before payments from federal programs get executed, what authorities would help the department identify and stop harsh payments for additional agency review?
Speaker 5
Thank you, Chair Comer. We are, as I mentioned, implementing a number of payment verification processes. So we're applying the technique of trust but verify, doing some basic checks before the agencies can certify a payment. One of the pieces that we are missing is the ability to ping authoritative federal databases to confirm a payee ID such as a tax identification number or a Social Security number.
Speaker 5
So we already received the data. We just can't verify it. So there are a number of, you know.
Speaker 4
Well, how does Treasury partner with states to prevent fraud in state-administered federal benefits programs? Are there ways that the treasury could increase or enhance their assistance? And I'm sure Minnesota wants no assistance based on what I've determined thus far in our investigation. But, how do you partner with states that want to work to prevent fraud?
Speaker 5
Thank you, Chair Comer. Do not pay, thanks to the Payment Integrity Improvement Act of 2019, authorizes the treasury to provide do not pay services to states that administer federal funds. However, it has been underutilized. We think that it can be part of a multilayered approach. So things like do not pay before the state issues payments to subrecipients would be very useful.
We can also work to address some of their common challenges by adding additional data. We know this works.
Speaker 4
And last question for mister Thomas. How can AI and machine learning be used to detect and prevent large-scale fraud schemes? What type of anomalies do these tools flag for investigators to follow up on?
Speaker 6
So, machine learn, I mean, all data science algorithms inclusive of machine learning and AI are gonna produce indicators of fraud. It's critically important, in each of the programs we've talked about today does this, has a fraud investigator and analyst who's an expert in the tools, techniques, and technologies that fraudsters use to look at the data coming out. So the types of things you're looking for are just as what were mentioned earlier, patterns of behavior that don't fit the expected patterns of behavior of someone who's using the money for the intended purpose or for the intended program design. We talk about at GAO, and we've published this, and we support the federal government, states, and local governments in using our fraud risk framework, which is designed to help them develop these indicators for a fraud risk management plan, which would then feed into algorithms, machine learning, AI, other data science methods, all acceptable, that could then be used to track and monitor potential fraud while the program's in execution. It's the purpose of it, as you design the tool, to find the behaviors that you want to get rid of.
Speaker 4
Very good. Well, we look forward to working with you all as we proceed with this fraud investigation that is starting in Minnesota, and I have a pretty good feeling that it's gonna expand to several more states. So with that, Mr. Chairman, I yield back.
Speaker 1
Thank you very much. The gentleman yields back his time. Mister Mfume, you're recognized.
Speaker 2
Thank you very much, Mr. Chairman. Miss Thomas, I wanna start with you because I just find some of what you said to be absolutely fascinating, almost unbelievable. And the main thing is the inability, and maybe Miss Kiskel, you can touch on this also, to be able to get tax IDs, social security numbers, or other identifying information that we've got in a number of different silos that you can't seem to get access to. The chairman and I both think that there's a way to break through this. We just need to know where the logjams are.
Speaker 6
Yes. So the data, yeah, data silos are a problem. I appreciate that. I've talked about it. You talked about it.
It's very clearly identified. You know, an example of the Social Security Administration with the death file is that they feel like the Privacy Act is making it difficult for them to share that with everybody. Now privacy is an important concept, particularly in the age of AI, but there are opportunities that the Congress could look into of modifying potentially a privacy act or putting in exceptions that allow the data to be shared specifically for the purpose of fraud investigations. That seems to be a fairly, you know, important component of some of the silos. Some of the other silos you've seen in our studies at GIO also had to do with incompatible or old networks, you know, modernizing the IT infrastructure of some agencies so that the data can be accessible to tools like DoNotPay as well as some of the analytic tools that a Prac uses.
You know, keep in mind that there are opportunities to improve the data and make it more available, the technology is old and does not have the ability to share information back and forth through APIs or other methods. That makes it very difficult. So those are some of the things we've talked about and made recommendations about.
Speaker 2
And what was the dollar amount saved as a result of the Social Security model? Was it 4,000,000,000?
Speaker 6
I think it's 4.7. Yeah. I mean, there's tremendous opportunity there to use this data together. You know? But it's a matter of, you know, these are organizations that operate separately, have their own infrastructure, and, you know, need funds to modernize and then make the sharing more available in addition to the statutory changes that we've talked about.
Speaker 2
And can you talk a little more about this digital services academy? It's a concept, I assume.
Speaker 2
For those of us who are not intimately aware of it or familiar with it, help me understand. I think this is about creating the people who will have the proper training to be able to do what's necessary.
Speaker 6
Yeah. So, you kinda think of, I like the chairman's idea of storytelling. That's kind of what is the ideal AI-enabled analyst look like? Well, it's somebody who's trained both in anti-fraud skills and investigations, but also knows a lot about data science, computer science, as well as artificial intelligence. And so in order to train those people, we need to set up an opportunity for them to get the training and leverage their already interest in being a public servant.
So the Digital Services Academy could provide that. Think about upskilling analysts that already have the interest and, you know, the passion to work in this anti-fraud domain, also just want some additional analytics, data science, and AI skills. That's the concept behind it, you know, help people develop this training. One of the challenges of this field, that we see in other technology fields, is that there's quite a bit of competition with the private sector. You know, they're obviously very interested in people with these skills as well for different purposes often, but, you know, that's a challenge for hiring people in the federal government.
Speaker 2
You talked about while we get to where we wanna go, that while we are where we are, that there are certain behaviors that sort of go outside of the norm that should call the attention to potential fraud that's occurring. I just need you to take a few moments to talk about what those behaviors are and why people who are administering programs don't understand that.
Speaker 3
Happy to take that. So what we're looking for again is patterns, trends, things that are unusual. And it all starts with a risk assessment similar to what doctor Thomas said about the GAO Fraudors Framework. When we talk about fraud, that's a broad term. So what specific concerns do you have, eligibility, identity verification, their financials, their claims.
And once we nail down what exactly we're most concerned about, we can determine what is expected, what's normal, what pattern. And then we can automatically, and machine learning can do this much better than a human being. Can say that the frequency of claims, the frequency of applications, is abnormal. That's three times what we would expect on a Saturday night. The frequency of claims from this geographic region, from this ZIP code, from this IP address is not anything we would ever expect.
And, again, a lot of this is based on statistics and data modeling. So it's not a human being sitting there saying, this looks unusual. It is literally a machine who pops it up and says, this is five times what is expected. And, again, once you start with the risks and then use the model, we can be much more efficient in addressing this issue.
Speaker 2
My time is that, mister chairman, if you would allow me one other question here. Are agency heads being told this? I mean, is there any training? If I'm the secretary of this or the administrator of that, I know I'm responsible for all the money. And if there's fraud, waste, and abuse, I'm gonna have to account for it.
But do they know what to look for in terms of abnormal behaviors, whether it's zip code or anything else that you mentioned?
Speaker 3
Sure. I can't speak to the entire federal government, but I can tell you I spent the last 29 years as a fraud investigator, mainly in grant fraud and public corruption. And there are agency officials at all levels that are intimately interested and aggressive about wanting to know from the OIGs. Every single OIG does an outreach program where they teach those willing to listen that these are the red flags, these are the concerns. But the biggest single challenge, sir, is that this is an agency responsibility.
And so they have to have commitment from the top to the bottom, tone of the top, to address the risk, identify the risk, address the risks, and seek out that. And we do see that in some places in the government, but it's not across the board.
Speaker 6
Yeah. And if I can just add, because at GAO, we do look across the government. And you're right. The mandatory requirement that we have recommended and we recommend to all agencies is they build an analytics-based fraud risk analysis team, so they do the work that my colleagues are talking about. And we make these recommendations.
We track these recommendations to make sure that, where they do respond to us, they implement them.
Speaker 1
We've now moved to a distinguished gentleman from Tennessee. Jones recognized.
Speaker 9
Thank you, Mister Chairman, ranking member. My mama used to have a saying, and I always hear her saying it in the back of my head. I hate to make my living off the suffering of others. And it seems that these people are ripping us off everywhere, and that poor folks can't get what they need because we've got limited money and unlimited needs. So I thank y'all for what you do.
I've always thought that those GAO reports, I've talked about those. It always reminds me of that last scene in Raiders of the Lost Ark where Harrison Ford just tells them, says, well, where is this? Where's the arc of the covenant? Oh, our top people are looking at it. Who is?
He said our top. And he said, well, and then you show up going into this vast warehouse. And I always think those GAO reports are there because, brother, I've been here for eight dadgum years, and haven't seen that dadgum report yet. But I'm fixing to, and I'll talk about that a little later, but not right now. What obstacles do y'all have the federal agencies face in this fraud prevention?
Either one and I'm both y'all, ma'am, sir or ma'am, your name's my name's Burchett, so nobody ever gets my name right. So I'm not even gonna tempt y'all. So thank y'all.
Speaker 3
Thanks for the question. Obstacles include access to data, but risk assessments, agency buy-in, resources. And I think the biggest obstacle is where to start sometimes. But I think, especially at the Prac, we've built some tools and have some great examples of where to start, and what some of the key lessons learned are from the pandemic. So I think the best starting point is what do we know about the risks that are out there, and how can we address this?
Ma'am?
Speaker 5
Thank you for the question. So I agree with Mr. Diffenbach. Treasury can play a unique role in helping to scale that prevention. We have a dedicated team of analysts, data scientists, and we can provide data more centrally. So not every little program has to create data connections to the important information that helps to safeguard the dollars in their programs.
So we can scale that prevention, and we can also help them understand best practices. And then, again, that last line of defense before the payments are made, being able to flag that risk.
Speaker 9
You're fixing to. I can guarantee it, brother. I mean, you're gonna be best friends.
Speaker 6
I love it. So, yeah, I mean, just building on what they've already said, we recommend a systemic risk management plan for all of these programs. And, again, these tools are great for leveraging by small programs, but the large programs as well should design the risk management framework we talk about into their risk management plan, and then build the indicators they need so the analytics will actually find the fraudsters. Again, before and avoiding the, we've talked about the pay and chase method before the money goes out the door.
Speaker 9
Y'all may have said this, but my limited billet comprehension is very limiting. I was when I was the county mayor, I found out that a lot of our different departments had different procurement avenues, and we consolidated those, and it saved us a heck of a lot of money. Of course, it cost some people their jobs, but it was just a duplication, and we needed to move that on. Do y'all think that this enhanced data sharing is really gonna help increase the detection and prevention of fraud or what just be wind addressing?
Speaker 3
Absolutely. I'll make one point that the federal government disperses a trillion dollars a year in federal grants. 85 of that, almost just under, over 900,000,000,000 goes to state governments to disperse, and they are absolutely a rich environment where we can leverage data to give them insights that they haven't had before.
Speaker 5
Thank you. Yes. We have lots of examples. So Do Not Pay also operates the back end of the Paris system, which helps states identify duplicate beneficiaries for programs like Medicaid. Last year, through that data matching, we prevented $1,300,000,000 in duplicate payments.
We're enhancing that with additional death data in November 19. States subscribed to that, and we identified 156 million in just that one month. So, yes, these things matter.
Speaker 6
Sorry? Yes. So that, over a trillion dollars you talk about that goes across all of the agencies and goes out to the states and localities, they're audited, and it goes into what's called the federal audit clearinghouse. Right? We built a tool that evaluates that and has identified patterns of indicators of risk, patterns of potential fraud, using machine learning and AI tools at the GAO.
It goes across all the different agencies of government. So, yes, all of these things matter. But, and you'll see this in one of our reports, we talk about the quality of those audits as part of the problem and validating that the auditors are actually doing their job and looking into the programs the way they're supposed to, so that that data does actually exist not only for that particular program, but across the rest of the government, so we can look at large patterns of fraud that do go across agencies.
Speaker 9
Thank you. I'm out of time. And, mister chairman, I think one of the things that we these good folks here do great work, and I think they're getting used just a little bit by the fact that we have too many staffers that, really, these committees are just large. There's no way that can be monitored. And I think that they are figuratively in bed with some of the lobbyists, and I believe they use these folks to do say, we're gonna do a report on this, and it takes you a year to get it done.
And then by the time you get it done, we moved on to another bright, shiny object. And both parties are guilty of that, and I think we need to get to the bottom of it. I think we need to utilize them more in a quick fashion. So thank you, Mr. Chairman. Thank you, ranking member, for y'all's indulgence.
Speaker 1
Chairman Burchett, thank you very much. And by the way, doctor Thomas, it'd be well worth your time to look at some of these important GAO reports and others that, inspector generals, because there are disconnects as you allude to. And I wanna thank the distinguished gentleman for his time.
Speaker 9
I'm on it like a cheap suit. Please contact my office, Mr. Thomas. Y'all, both y'all, all three of y'all. Thank y'all.
Speaker 8
Thank you, Mr. Chairman. Thank you to our witnesses for being here today to discuss substantive reforms to address this important issue. As my colleagues noted, we on the government operations subcommittee have been focused on fraud and improper payments, not as a partisan issue, but as one that must be addressed for the benefit of all Americans. We have heard in today's testimony about the importance of catching fraud before program benefits are paid, rather than tracing down fraudsters to try to recover funds after the fact. I support efforts to remove agencies, regard agencies toward this more effective strategy, which will also leave a greater portion of funds Congress appropriates available to provide services and support to intended recipients.
It is also critical that agencies conduct their fraud prevention activities as efficiently so that funding and services can reach the people without unnecessary delays. Miss Musk Miskol, what are some ways the federal treasury department is trying to help agencies conduct fraud prevention checks today more quickly?
Speaker 5
Thank you, Chairwoman. So I mentioned do not pay. We are making it more useful by adding additional databases. So focusing right now on commercial databases, and with more authority, those really high value federal databases. So we want every program, and there are thousands of federal programs, to use Do Not Pay before making award and eligibility determinations.
We also have a concept called continuous monitoring. So you may say that an organization looks like it passes all of the checks, but something may happen, like found in an audit. This continuous monitoring can flag that risk, you know, as payments are going out the door. And then finally, treasury as that, central disbursement organization can flag risk. We have a unique role where we can see cross-government payments.
So when my friend here, Mister Diffenbach, tells me that a specific entity or individual is at risk, we can scan the cross-government payments and identify other programs at risk.
Speaker 8
Mister Thompson, what tools and resources will agencies need to be able to effectively prevent fraud without unnecessary delays in payments?
Speaker 6
So they need both the fraud indicators for the programs that they have set up, but also the data science tools that we've been talking about that the prac has been using, as well as access to tools like DoNotPay. All of these are examples of tools that can be implemented in a fast way, so that you can prevent potential fraud before the money leaves. Now we have to set them up so they move quickly and have access to the appropriate data for each of the programs. But that's part of what the fraud risk framework we provide is all about.
Speaker 8
I hope this subcommittee will continue to focus on substantive reforms rather than partisan attacks, and I yield back.
Speaker 1
Thank you. I appreciate the gentleman. Gentleman gentlewoman yields backward time. Now I would recognize distinguished young chairman from Alabama, Chairman Palmer.
Speaker 10
Thank you, Mr. Chairman. I've been working on reducing our improper payments for 11 years since I've been in congress. One of the first things I did when I was on the budget committee and insisted that we take reducing improper payments into account in our budget process. And I'm sad to say that, when I came into Congress, our improper payments were running somewhere around $130,150,000,000,000 a year. It has since surpassed that considerably.
During COVID, it got even worse, particularly with payroll subsidies that we sent out. The additional unemployment insurance, the fraud was massive on that end. I think one state in particular, in the first four months, sent out about a billion dollars in fraudulent payments. I know that, we're working on that, but I wanna know what progress has been made in that area to address the fraud from the pandemic funding.
Speaker 3
Thank you, Congressman Paul Mart. We've done a tremendous body of work on identifying what went wrong during the pandemic. We've issued lessons learned reports and a blueprint for program integrity about how to ensure this never happens again. But let me answer your question quickly with a story that chair sessions, you know, will appreciate. One of the many projects we did was we examined recipients of HUD low-income housing benefits with the Social Security numbers that were also used to obtain Small Business Administration PPP loans.
So folks that were claiming a low income to get the housing benefit and a high income to get the PPP loans, often forgivable loans. We found 40,000 instances in which the disparity in income between those two programs was 10 times or greater, and that impacted $860,000,000 in PBP loans. So I think telling those stories is important. That's what one of many risks and programs. And the end to that story is that ID theft or lies occurred in the HUD program or the SBA program or both, and legitimate victims that needed housing or PPP loans didn't get them.
Speaker 10
That's been mentioned a couple of times, but one of the problems with all of this is the failure to have data systems that interface. Correct. We, I was astonished at the fact that we couldn't even get agreement within Congress on access to the Social Security master death file so that we could check that against some of those things that were submitted by the states. I also found the g o GAO was tremendously helpful, Mr. Thomas, in this. But about 55% of the problem was an administrative error.
So I heard to verify eligibility, which I think is a massive problem right now, and antiquated data systems. And I'd suggested that one of the ways to approach this to make major improvements in agencies' ability to eliminate improper payments is that rather than be punitive toward the agencies, we say that whatever you save, we will take part of that to bring your data systems in the twenty-first century so that we have the ability to interface across the entire federal government. Would you like to comment on that, Mr. Thomas?
Speaker 6
Yeah. I mean, certainly, changes in incentives are something that we have recommended. I do just want to add a note of, not caution, but just thought is that improper payments are not always recoverable. They could also just be, I understand. Yeah.
They could just be errors in the dataset. So when.
Speaker 10
You got over 200,000,000,000, if you were, I mean, if you recovered a tenth of it.
Speaker 10
That would pay for a lot of what we need to do in terms of improving our data systems.
Speaker 10
The other thing is that, and I know that, some of this is a sore subject with some of my colleagues, but some of the bills that have been passed left some pretty big doors open. And the Inflation Reduction Act, with some of the green energy subsidies, the ACA premiums were, I think the GAA report was billions of dollars in fraud on the tax subsidy for the ACA. Same thing was true then in the infrastructure and Jobs Act, improvement in Jobs Act. I think that Congress has got to do a better job of writing legislation that helps that, but we're really gonna be dependent on federal agencies to do the proper oversight here. And then when you need help, you've got to come to us.
Because this is a massive problem. When you're talking $200,000,000,000, you're talking 2,000,000,000,000 over 10 plus interest. We're borrowing that money to send it out. Thank you, Mr. Chairman. I yield back.
Speaker 1
Thank you very much. I appreciate the distinguished gentleman. His avenues of working in this area for his career have been not just stellar, but they have really illuminated many people, including conversations that you've had with our team about the need to make sure that they see the bigger picture also, and I wanna thank the distinguished gentleman. Gentleman yields back his time. Mister Frost, you're now recognized.
Speaker 7
Thank you so much, Mr. Chairman. I'm glad we're holding this hearing today. I support innovation to tackle fraud that's executed safely. And we've heard testimony today about the promise of artificial intelligence and other new technological tools that may be used to improve fraud detection across the entire federal government. That said, machine learning and artificial intelligence systems leverage massive amounts of data, and specifically personal data.
Miss Miskol, how does the treasury protect data, and how can I address concerns from my constituents on their data privacy?
Speaker 5
Thank you, Congressman Frost. I sincerely appreciate that concern. We take privacy security very seriously at the Treasury Department, specifically within Do Not Pay. Privacy is built in by design. We follow the principle of least privilege, meaning that a person can only receive a response back on the information that they provided.
We operate in a FSMA-high environment, which means that it's the highest standard in terms of federal cyber. We apply continuous monitoring, and we are also transparent to the public. We provide transparent systems of record notices, communicating what data we have on individuals and how it can be used.
Speaker 7
Thank you. Mister Thomas, what are the concerns about having unknown quantities of private data and sensitive information in a master database?
Speaker 6
So, you know, it's kind of the there are technologies that if you have them in separate data systems, that you can control the way they query back and forth to keep the data private and controlled. But you're right. Once you put it in a master system where everything is all together, which is the design of AI, AI needs that to train. All of a sudden, you're now dependent on the actual AI prompt to control your privacy. The term that the AI industry uses for that is alignment.
There's actually been excellent studies recently in how you can actually break alignment in an AI system so you can get it to return information back to you that is not designed to return to you, and even safeguards were put in. So this is an area of concern, you know, and that's why we talk about it's important to use potentially the least complex solution to trying to get to where we are using machine learning where you don't have to train on everything, but you can still maintain the controls that treasury is currently doing with their do not pay system.
Speaker 7
Yeah. Thank you. I know there's, you know, many different things that are being done to make sure data's kept safe when people query information or when information goes from agency to agency. But part of the reason I have this concern is because it's also dependent on the conduct of the administration and different things that are going on. So, for instance, you know, under President Trump and back when Elon Musk was more involved than Doge was going on, they conducted attacks on our federal data systems and cybersecurity.
Right? And this is the public information. Right? They brought unsecured servers into the federal agencies. They reportedly exported unknown quantities of private data that we still don't really sure how much it was or where it was from.
And then the one that's most concerning for me is they contracted with private companies to merge American sensitive information into a master database that would give the administration surveillance powers and put data at greater risk of hacks from outside people. And so it's just important to me that we ensure that any new innovative tools are not politicized and misused by this administration and any administration, and that they aren't abused by private corporations that have been brought in to help us with different projects. We also have to recognize and expand successful existing tools, and not toss them aside either. There are tools that some agencies are not using enough that I think we need to be explored. So, miss Cole, the treasuries do not pay system, which you brought up earlier, is one such effective program.
What do you hope other agencies can learn from this model?
Speaker 5
Thank you, Congressman. Agencies can learn a lot from using DoNotPay. We are helping to integrate the data for them. So we talked a lot about, you know, technical debt, long-standing issues with resources. We can provide that information securely to the people that need it.
And I just wanted to mention in the conversation around AI, we have successfully used machine learning to detect check fraud. It's been an outstanding use case. In 2024 and 2025, because of this technology, we prevented and recovered $1,900,000,000. So, again, used responsibly, this technology can be really transformational. But as Mister Thomas mentioned, it relies on the data.
So we had really good data about the checks we issue, and we could compare that. And that's what's important about using AI.
Speaker 7
Yeah. And, obviously, more and more interagency collaboration is important too, and is free to use. Last question, real quick. Mister Thomas, what basic fraud protections could agencies use that they aren't already using?
Speaker 6
Well, as we talk about, you know, implementing our fraud risk framework, and developing models as programs are being developed is one of the most important things they can do, in addition to actually implementing data science-based fraud programs. So they are bringing in data scientists, implementing the algorithms that the prac has been able to implement, proven tools that can identify fraud that, just like you said, exist today.
Speaker 7
Thank you. And thank you for letting me go over, Mr. Chair. I yield back.
Speaker 1
Thank you very much, Mr. Frost. And we appreciate you being at this hearing today. Your support of this is very important. And you have been to all of our hearings and not only showed up to ask important leading-edge questions, but I think worked as part of the team, and I appreciate and respect that very much. Okay.
We've got several members that Aspie waved on and others who have indicated they might be here. I've chosen to put myself as last, but pending them arriving, I'm gonna go ahead and use my time. All three of you have been very good, I think, rather exceptional at taking me up on this view about telling the story. And I think mister Infume really began some of this with mister, for sure mister Deffenbacher. But I wanna go to Mister Thomas, Doctor Thomas if I can first, and then have either of you then join in.
This telling the story really, that is the power of AI, not just machine learning, but across the avenues that each of you have. And I'd like for you, doctor Thomas, to not recreate what we talked about yesterday in my office, but pretty close to that about how important these exercises are on a really program by program basis that you have found characteristics that you spoke about, about how certain things that might be a request under one program found themselves across, if you used AI to highlight where there were inconsistencies, to look at inconsistencies that would draw you to those things. Do you mind taking a minute? Mister Infume, I think, would learn a lot from this. Perhaps he knows it, but hearing from you, Dr. Thompson, would be important.
Speaker 6
Yeah. I appreciate that. So we have a demonstration program that I talked about earlier that we call FACET that is really built for, just as chairman Sessions talked about, identifying indicators of potential bad behavior, really just data points in a large data collection. And this is the federal audit clearinghouse that, you know, can be used across programs. So I'll give you a couple of thoughts there.
The federal audit clearinghouse, you know, houses over 1 trillion dollars of spending every year. And these are audits that are done not only on federal programs, but these are state programs that we're using federal money for and local programs we're using federal money. We use a combination of AI, natural language processing, and machine learning to identify, you know, one pattern here, say, that, you know, is in a food support program that then is replicated in another program, completely different in a completely different locality. These indicators of fraud don't mean fraud is happening, but it means someone should take a look at it. And that's really the message we try to put people is that not only should they look at it, they should understand, is it actually a problem with the program?
Is it a problem with the way the audit was done? Or was it a problem where the design of the program isn't actually collecting the right piece of data? And then you can help that program say, okay. All new grantees should produce this piece of data. We'll then go into your fraud risk model.
And it's, you know, we've been doing it now for several years, and it's produced lots of wonderful examples. We actually published several of these examples on a website, that is basically a pattern storage site of the patterns of fraud that you can go to at the GAO's website and look at and see these different patterns that have been identified through this tool and other tools as well as some of my colleagues' tools.
Speaker 2
Could you make that available to the committee, this publication? Yes. It is a publication. Right? I'd really like to see it because I'd like to see what are the examples that.
Are so clear that people are missing.
Speaker 6
Yeah. We'll provide the reports, and it's a live website you can go to. Yeah. Thank you.
Speaker 1
Miss, anyone else have once again, this is telling the story about how important it is, and it might be Social Security, maybe a death file, it may be something else. And I know that we could say, well, there's hesitancy by the agency. Congress, you need to address that. I get that. But I'm more into the power that really exists of the tools that you have if we allow you to fully utilize them.
If we look at them and say to you, yes, please go do this. So I'll do, anybody wanna add to that? Otherwise, I wanna go to my last question. Mister Jack is here, and he's taking time, and I wanna get to him.
Speaker 3
Thank you, Chair Sessions. Great question. Under the umbrella of artificial intelligence, we are leveraging it. We have built the fraud prevention engine, as I talked about, that uses a variety of techniques. And the best analogy I can give you is if you put the best government analyst in a room and give them an application for federal funding and access to databases, they can check and they can review and they can look at auto reports and they can look at prior history and all kinds of things.
This machine can do that very rapidly, almost instantaneously. The system we built can do 20,000 applications per second, and it's using all kinds of those technologies to give us an edge on this information war, if you will, to shine light on what's actually occurring. That's, to me, the single biggest challenge: fraudsters are hiding what they're doing, but using these technologies in a responsible way with the right datasets, addressing the right risk, can give decision makers instantaneous visibility into what is actually occurring.
Speaker 1
Yes. One last question because I know I'm at my time frame too. I wanna ask you. There's a great conversation last week and also in the media across the country, and that deals with the states. And, we've already heard, Miss Gold, talked about how this is a huge amount of money.
Tell me about your recommendation to mister Nfume, myself, mister Jack, and this subcommittee about how we ought to be looking at conversation with states. For instance, we know that some of the states want to do business with us, some don't. We know that, however, we have federal money that is at risk, and I think that they would, through our efforts, want to do things. Why do we give them money? Do we ask that they update their databases?
Do we share information? There may be 10 or 15 things that I could think of, but you are closer to that than I am. What would be your advice to this subcommittee about doing business with states on data and information related to federal programs?
Speaker 3
Sure. I've done fraud investigations over my career in, I think, 35 different states. And I can tell you that they vary dramatically in their interest, their resources, their level of responsibility. So I think it starts with conversations with those states that are interested in starting the conversation as a pilot, as we discussed in your office yesterday. I think our tools we built to the prac, along with the geo risk framework and other tools from treasury, we can add tremendous value to what they're doing.
Again, I would submit it as a pilot project to prove its worth. But with additional insights to those states in particular programs, looking at particular types of risks that they can't wrap their arms around, they don't have the data for, and build off from that and prove that the concept indeed can work, which I think it can.
Speaker 1
Members of the panel here, we were meeting individually, separately, and so they did not have the advantage. And, I asked where would you start and ask the states, and then I had my ideas. And I really think that it's a conversation that we wanna have, mister Infume, on and each of you and perhaps some inspector generals and perhaps some in your agency. So I see that as a definite, walking out of here first thing we need to do in addition to the legislation that I think mister Infume and his members in mind would agree to. I wanna thank you very much.
I'm sorry, I'm well over my time. We will now move to the distinguished gentleman, Mister Jack Jones recognized.
Speaker 11
Thank you very much, Mr. Chairman. I wanna thank our witnesses for testifying today. And at the outset, I always like to frame this for constituents and everyone watching back home. But one of the things I wanna focus on today, and the chairman's probably too humble to acknowledge it, but one of the accomplishments that he had thus far in this session, among many others, over his career, was the inclusion of HR 2277, the FACT Act, and the One Big Beautiful Bill Act, which, of course, as mister Giffenbach no doubt knows, extends pracs authority through 2034. And I think it's important to focus on that at the outset because we talk about the many accomplishments of the One Big Beautiful Bill Act.
This is one of them. And it's because the bill was so big and had so many great things within it that sometimes some of these accomplishments could be overlooked by the media and others. But in this case, this hearing amplifies just how successful that was. So, wanna commend you, Chairman Sessions, for your leadership in introducing that legislation and getting it into One Big Beautiful Bill Act for passage and enactment. So my first question to you, mister Tipfenbach, is to help our constituents know my constituents and the constituents of this committee, could you walk us through what the world would look like if Prac had expired?
If we no longer had access to this great, you know, great tool, that is meant to root out waste, fraud, abuse. Walk us through what life would look like if this had expired and we had not been able to extend it last year.
Speaker 3
So thanks for that great question, congressman Jack, and for your support. The Prac has assembled a phenomenally unique set of data about pandemic fraud, about program fraud in general, about the patterns, the trends, the anomalies. We've issued a number of alerts. So had we expired, the ability to provide the insights I just spoke to a minute ago would be gone. There would be another disaster.
The congress would fund an emergency data analytics capability, and we would have to spend a year or two to rebuild that. So we've been able to continue to keep pace. The fraudsters don't take naps or breaks. So we've been able to continue to build upon everything we've learned over the last six years, and I think it's been a tremendous asset to the congress and to the taxpayer.
Speaker 11
Well, thank you very much for your answer. And I'll next move to Miss Miskell. I'd first like to address it if I could. I'm very interested in the Department of Treasury's do not pay system. Frankly, I've got a couple companies in my district that work on this.
So I'm uniquely interested in the topic. And I'm just curious if you could share with us some of the recommendations you may have, legislative action you'd like to see that would help make the system more comprehensive and even more effective than it already is. Welcome your thoughts in that regard.
Speaker 5
Thank you, congressman. And to just put it simply, how do we best equip states? How do we best equip agencies to prevent and detect fraud? It's making it easy to do the right thing, and we can do that with data. So do not pay can be that simple tool, but it's only as effective as the data within it.
So there are a couple of key data sources that are federal data sources that we have had a tremendously difficult time accessing, that we know would be very valuable from GAO reports, from PROC recommendations. Those datasets include the ability to verify financial status. And I'm not saying, you know, exact numbers. I'm saying a threshold like, does mister, you know, so and so make over 1 million dollars last year? Other data sources would help to verify identity attributes.
Again, the taxpayer identification number, the Social Security number. And, again, we're not looking to get full access to these databases, but just a simple yes or no to identify identity theft. There are a number of commercial databases that we have found to be extremely valuable as well, and we are working to rapidly onboard those. So it's not just federal. It's also commercial.
Speaker 11
Wonderful. Well, I hope you've had a chance to get to know my former colleague, Francis Brooke, who serves in treasury alongside you. He's a wonderful guy. So I appreciate your testimony here today. I will close with mister Thomas if I could.
Right as I was listening a few minutes ago, you were talking about AI and the development of it. I'm fascinated by the utility that could be offered in this space. Welcome. Any closing thoughts from you on how best AI could continue to strengthen our mission to eliminate waste, fraud, and abuse?
Speaker 6
Yeah. There's, like I said, there's tremendous opportunity in AI. I think the foundational components that Ashley Miss Gold was talking about are critically important, and that is build a solid data collection of what is AI or what is fraud so that AI can learn that. The challenge with AI is that it doesn't know the difference if you don't tell it. And so if we properly label all of these examples that both of my colleagues are finding, collect the data that represents this is fraud, we can then start to train a tool to do this broadly.
And then all of a sudden, you can leverage without having to have the analyst read all of these audits and go through the data by themselves. They can be empowered with this tool that can pull out these indicators of fraud. These are the patterns that just don't represent typical behavior of a payee in this program, and now someone can go look at it. That is only possible if we have this foundational kind of gold standard of this fraud database.
Speaker 11
Well, thank you all for your testimony. I learned a lot from today's hearing, and I yield back to the chairman.
Speaker 1
Gentlemen, yields back. Mister Jack, thank you for taking time. I know that you were stressed in your duties today, and I appreciate you taking time for this important event. I have been asked to ask this question, so I'm gonna ask myself if I could. Mhmm.
You know, since I've agreed to. Doctor Thomas, which agency manages the Federal Audit Clearinghouse?
Speaker 6
So the Federal Audit Clearinghouse is actually the responsibility of the Office of Management and Budget. They delegate that to the General Services Administration, and that is actually who's operating it. And when we access it, we work with them to work through their APIs. Now they've had some challenges with staffing recently, which has delayed our ability to make some updates. I'll give you a specific example.
Several of their newer APIs were mislabeled on what versions they were, so it made the system not work appropriately. But that is, you know, something that could be improved, and so that's where it lives.
Speaker 1
So what would you say is the current status recognizing, as you have alluded to, the delay and perhaps the misrepresentation of the data. Where does that stand today, and who has the responsibility on your side at GAO and, at the, within the agency?
Speaker 6
So, the agency doing the work is the GSA. They're getting the responsibilities at OMB. Within our group, it is our innovation lab that is working with them to identify these errors and improve them. Now, we have recently been working to get better response from them, but I think it gets back to my earlier statement of, you know, just enabling them with a, you know, a workforce that has the technology skills to run an important program like this. And there's over a trillion dollars a year in spending in the federal audit clearinghouse.
It's a rich dataset if we work to improve it and extract value from it.
Speaker 1
So that's what's at risk, that trillion. I'm not trying to give the whole example of a trillion, but that is how big of a problem this is would help solve.
Speaker 6
Much is obligated in the federal audit clearinghouse database. Yeah. Those, that's what the audits represent.
Speaker 1
Well, I respect that, and we will take that up also. I think that this subcommittee is very capable of looking at things. And if a trillion dollars is not big enough, we're in the wrong business. Mister Infume, do you have any additional questions that you would like to engage this panel on?
Speaker 2
I don't know if I have any questions. I just have a couple of observations. You know, you and I have been at this for some time now, and, fortunately, gotten any gray hair from it. You just can't see it. Oh.
Fortunately and unfortunately, this has gained a new set of energy and a whole new set of interest because of recent events in Minnesota and elsewhere around the country. I just think that we're on the precipice of something really big here if we do it the right way. And, you know, to the extent that we can get some interest going in the other body, meaning the Senate, and find a way to take advantage of all the information we've got to be able to convey it to other members of the House of Representatives so that they get a burning desire to recognize that whether we're talking about the do not pay mechanisms or cross sharing of information with these large datasets or mandating agency heads in a particular type of responsibility, The Digital Service Academy to train, and even though there's a great deal of competition on the private side, behavior identification. I just think that Prac, OMB, the treasury GAO represents a great deal of information that we should have the ability to try to coalesce and put into one great big effort. And I would think, mister chairman, just as we look back over the history of this congress, things like Sarbanes-Oxley, which changed the landscape in many respects, there is an opportunity here.
I'm just having a little difficulty identifying. I feel it, but I think the more we work through this, we'll be able to identify it. And so I would ask the witnesses, please, to commend to our attention, and particularly mine. Any journals, any writings that are current and recent, your own thoughts, white papers, anything that has in it a set or subset of information that you think any of us who are looking at this could benefit from. And in our own way, mister chairman, we're gonna have to find some sort of way to absorb it all.
It's a lot. But I just see that it's like begging for some sort of national response that would be so strong that it would set the course over the next five or 10 years. And at the rate AI is moving, we're already behind, in my opinion. So, I think this is a good hearing. Clearly, I'm glad the witnesses have come out.
I thank you for your support of this, and I would yield back.
Speaker 1
Distinguished gentleman yields back. I would like to go back to a conversation that mister Jack really had with us talking about some of those things that you talk about successes and the things that we're making sure that Prague did live on. And mister Jack, congressman Jack had it right, but we also did it together. We did that together, and I think that if we view the assets and resources of the organizations that are here today and really take them at their word and work with them. You and I work together.
This subcommittee, I think, can really make a difference, and perhaps we need to involve the other body a little more. But I think that we've been successful at putting things in a must pass piece of legislation that would make things better. Too right. Two things. Number one, I'd ask, and now recognize myself for a UC request.
I would like to enter into the record two letters, one from the Program Integrity Alliance and the other from the United Council on Welfare Fraud. And these were on reforms needed to prevent fraud. They're their observations. Without objection, I would enter those into the record. The second thing I'd like to carefully address is some conversation that was had today in relation to staff.
The staff of the subcommittee, the staff of the committee. And I want to commend our staff members on what I believe are professional products for the right reason, in the right way, with the right outcomes. And I don't have to engage in a yelling match on this, but I want to say to each of my staff members, thank you. Thank you to their service. Thank you for making sure that they stand up in a professional manner.
Many of these people work long hours and do things that might take a lot of time, but I want to reiterate that my opinion, and I only have 27 years at this, few times have I seen a staff that I felt like would have been engaged in anything that would be, unbecoming to professional conduct. Secondly, I want to acknowledge, and we do this often, because Mister Mfume and I have oversight over federal government operations. All federal government operations, we have original jurisdiction too. I think the federal government has great people also. I think the federal government has people who are devoted and dedicated.
They do their job. They are public servants. They get up. They do things that match or try to match the needs of the American people. In every organization as large as the government is, I'm sure there will be some that are through their time, and they're ready to leave or ready to do other things.
That's fine. That's fair. But I think that I've seen a devotion and dedication across government. And I think mister Infume and I, from our vantage point, would say that we have confidence that we can move to a brighter, better world and work with AI, and it will help us in so many ways. Does the gentleman wish to make any closing statement?
Speaker 2
Well, just what I hope and pray is an obvious sort of reality. I keep going back to that dollar amount of $533,000,000,000 as a max and 233 as a floor. That is a lot of money. If we were able to just half that, I mean, think of all the things that we could do in this country. We wouldn't have had a situation where SNAP benefits were being held from children all across the country.
We wouldn't be in a situation where we couldn't figure out how we're gonna provide ACA credits for people who are about to see gigantic increases in their health care costs. There are so many more dollars we could be sending to states and municipalities. So it's like we're almost beggars sitting on bags of gold, and that is why, again, finding a way to get our arm around this issue, and to take that money that's going out the door that the criminals and the bad guys have been taking for years, and to keep it in house so that we can do some good with it. If that does not compel, mister chairman, our colleagues who want to be a part of this, I don't know what will. So I'm gonna be optimistic and hopeful at this point that that's such a compelling argument that you and I won't have to sit here for another 3 or 4 years trying to make this case, waiting for others to catch up.
I'll yield back.
Speaker 1
Mister Mfume, thank you very much. So let me address the three of you. You've been impressive. You are singularly impressive, and your ability to know each other and work together on a common issue is important to the country. It is always hard to know as you're working on something where it's going to end up.
But I think that by us working together, we will speed that effort to get where we wanna go. One thing you can get from mister Infume, as we promised last year and the year before, we were gonna follow up. We were gonna do the things that would be necessary to empower the good outcome. We were going to make sure that we ask questions that you consider to be thoughtful and something that would lead to a better outcome. I want each of you to know how much we appreciate you, how much we recognize that if we take your nuggets of data and information and good ideas about a better outcome, we can get there.
And so, mister Mfume and I, just a second, are gonna come down and shake your hand and thank you for being here. I hope you're able to go home tonight and say to those that you love the most, perhaps that love you the most, that you did a good job today, that you presented yourself in a professional way, that you defended the work that you do, and that you're asking for more of it, not less of it. With that said, without objection, all members have five legislative days to submit materials and additional written questions for the witnesses, which will be forwarded to the witnesses. And if there's no further business, without objection, the subcommittee stands adjourned.
Ask anything about this video — I analyse its transcript to answer.