Skip to playerSkip to main contentSkip to footer
Forbes Breaking News
  • 2 days ago
On Wednesday, the Joint Economic Committee held a briefing on DOGE and AI in government efficiency.

Fuel your success with Forbes. Gain unlimited access to premium journalism, including breaking news, groundbreaking in-depth reported stories, daily digests and more. Plus, members get a front-row seat at members-only events with leading thinkers and doers, access to premium video that can help you get ahead, an ad-light experience, early access to select products including NFT drops and more:

https://account.forbes.com/membership/?utm_source=youtube&utm_medium=display&utm_campaign=growth_non-sub_paid_subscribe_ytdescript


Stay Connected
Forbes on Facebook: http://fb.com/forbes
Forbes Video on Twitter: http://www.twitter.com/forbes
Forbes Video on Instagram: http://instagram.com/forbes
More From Forbes: http://forbes.com
Transcript
00:00:00Opening statement. Okay, let's find the opening statement.
00:00:05First up, I want to thank all of you.
00:00:07Apparently, you've all done just a terrific job working with our different staffs in educating us.
00:00:13You've seen our idiosyncrasies of wanting you to fixate on the things where we can actually make some policy changes that actually make a difference.
00:00:20So, look, for six-plus years, because of the quickness of our office, we've been introducing AI bills, audit the Pentagon, using technology, using a crawler to find out where assets are.
00:00:36Health care. If over the next decade, interest in health care are the primary drivers of U.S. sovereign debt, could we actually still protect the quality of health care, maybe even make it better,
00:00:47but using technology to deal with those cost drivers.
00:00:52And I know sometimes that makes people very uncomfortable, but there was also a time that going to Blockbuster Video was the family ritual on Fridays, and today you hit a button.
00:01:03So, now I'd like to introduce our distinguished witnesses.
00:01:08You all get to be distinguished.
00:01:09Dr. Sterling Thomas is the chief scientist at the Government Accountability Office.
00:01:14He leads the work conducted by GAO scientists, technology assessment, and analytics team, which provides information to Congress on emerging technologies and their application in government.
00:01:27Prior to his appointment at the GAO, Dr. Thomas served as chief scientist at Nobelis, a nonprofit research institute,
00:01:37where he applied machine learning and artificial intelligence to his research on synthetic biology.
00:01:46Okay, you already realize you just like doubled the IQ of, never mind.
00:01:50Dr. Brian Miller is a practicing hospitalist and professor of medicine and business at Johns Hopkins,
00:01:58and somewhat of a friend.
00:02:00Dr. Miller is also a non-resident fellow at the American Enterprise Institute, where he researches and focuses on healthcare competition,
00:02:11FDA policy, public health, and integration of AI into the healthcare system.
00:02:18The doctor is head of AI policy at the Abundance Institute.
00:02:24His research focuses on how artificial intelligence can be used to maximize innovation and advance human interest.
00:02:40Previously, he was the chief technologist at the Federal Trade Commission,
00:02:45where his work focused on economics of data privacy and blockchain.
00:02:49We should talk.
00:02:51I worked on blockchain before coming to this.
00:02:54Now, I'd like to turn over to the ranking member to introduce the Democrat witness.
00:02:59Well, thank you so much, Mr. Chairman.
00:03:03And after I introduce the witness, I have a couple of opening remarks, if that makes sense.
00:03:07Why don't you go straight to opening remarks, because you may get stressed with having to vote on it.
00:03:10Okay, and then I will introduce Mr. Canarsa.
00:03:13So, I just really want to thank you for calling today's hearing on a really important topic.
00:03:18And to your point, a topic that has a lot of bipartisan interest and a lot of bipartisan potential.
00:03:26As this is the first JEC hearing this Congress, I also want to say how much I look forward to working with you,
00:03:31Mr. Chairman, in your new role as chair of the committee and my new role as ranking member.
00:03:36I really look forward to working together, and I've appreciated our conversations so far.
00:03:40And I'm glad we're starting out with a topic that we can find some real common ground on,
00:03:44cutting waste, fraud, and abuse through the use of innovation.
00:03:48I also want to thank the four witnesses for testifying before the committee and sharing your expertise on this topic.
00:03:54We need to keep working together across the aisle to save taxpayer dollars by reducing waste, fraud, and abuse.
00:04:01As we will discuss in today's hearing, there are many ways to do so by embracing new technologies and better data analytics.
00:04:07These types of tools can help an investigator identify potential fraud or a caseworker streamline an application process.
00:04:15I know that all four of our witnesses today will outline ways that the government can better deploy technology.
00:04:20I really look forward to building off these ideas moving forward.
00:04:23Before we get to the introduction of our witness, though, I do want to just note my serious concerns with the reckless action so far of Elon Musk and Doge.
00:04:35Whether the government is providing health care for veterans, social security checks for seniors, or loans for small businesses,
00:04:41we should continually strive to improve the taxpayer experience.
00:04:45In addition, I want to bring the committee's attention to the fact that the administration's decision to fire 17 inspectors general who investigate fraud,
00:05:00that's what they do, is completely counterproductive to the goal of improving government efficiency.
00:05:06I firmly believe that we should be cutting waste, fraud, and abuse,
00:05:09and that we can do so without making life harder for children, families, and seniors all across our country.
00:05:14I hope the conversations, including the one we will have this afternoon, can help us chart a better path forward.
00:05:21So, again, I thank you, Mr. Chair, and to our witnesses for agreeing to testify today.
00:05:27And happy to proceed with the introduction.
00:05:31Mr. Andrew Knarsa is the Executive Director of the Council of the Inspectors General on Integrity and Efficiency.
00:05:37That's an independent part of the executive branch focused on supporting the work of inspectors general.
00:05:43Prior to his current role, Mr. Knarsa served in the Social Security Administration Office of the Inspector General from 2009 to 2023.
00:05:51Welcome, Mr. Knarsa.
00:05:54I really look forward to your testimony as I do the testimony of all our witnesses.
00:05:57Thank you, Senator.
00:05:58And please don't be discombobulated if you see members popping up and popping down.
00:06:03It is the nature of being here during a fairly chaotic time.
00:06:06Dr. Thomas, please, five minutes, educate us.
00:06:10All right.
00:06:11Chairman Schweikert, Ranking Member Hassan, and members of the committee, thank you for inviting me to participate today.
00:06:18In today's hearing, we discuss how AI can be used to reduce fraud and improper – waste fraud and improper payments.
00:06:24As GAO's chief scientist and throughout my career in industry and academia and in the intelligence community, I've seen great advancements in data science.
00:06:33These innovations offer exciting opportunities to improve government efficiency.
00:06:37As you know, GAO is the nonpartisan watchdog for Congress.
00:06:41We have expanded our science technology team in recent years, and it includes a team of data scientists.
00:06:47We know firsthand from their work that AI holds great promise in furthering GAO's mission and your goal of safeguarding taxpayer dollars.
00:06:55My aim today is to offer three important actions that will help us reach that goal.
00:07:00First, we must continue and augment our traditional anti-fraud efforts.
00:07:05Second, we must apply AI thoughtfully and ensure that we use quality data to mitigate its well-known risks.
00:07:13And third, we must ensure the federal workforce has the skills they need to apply new innovations like AI.
00:07:20Regarding my first point, GAO has a large body of work on fraud and improper payments in the federal government.
00:07:26Last month, one of my colleagues testified before House Appropriations on this topic.
00:07:32The federal government reported an estimated $162 billion in payment errors or improper payments during fiscal year 2024,
00:07:40and that is almost certainly an underestimate.
00:07:44We have outlined numerous ways that Congress and federal agencies can tackle this problem with existing capabilities.
00:07:50For example, Congress could make permanent the Social Security Administration's authority to share its death list with Treasury's do not pay system.
00:07:58These traditional methods are routinely preventing and detecting fraud.
00:08:02By enhancing them, we can save taxpayer dollars today without any new technology.
00:08:08But on my second point, I am optimistic about innovation, but we must be thoughtful about it.
00:08:14In data science, we often say garbage in, garbage out.
00:08:18Nowhere is that more true than with AI and machine learning.
00:08:21If we start trying to identify fraud and improper payments with so-so data, we're going to get so-so results.
00:08:28Broad use AI is still at early stages of development and implementation, and rapid deployment without thoughtful design has led already to unintended consequences.
00:08:38Before pouring data science on the problem, we need solid, reliable, ground truth data and a human in the loop to ensure data reliability and application of this technology.
00:08:49GAO has an AI accountability framework which lays out these and other principles.
00:08:54One piece of advice that emerges from such principles is to find a solution that produces the desired result with the least complexity.
00:09:02For example, in response to our recommendation, the Small Business Administration screened all PPP or Paycheck Protection Program loans made before September 20 with a rules-based tool.
00:09:14It was just looking for indicators.
00:09:16It wasn't using AI, and they still identified $4.7 billion in loans that went to ineligible recipients or were used for unauthorized purposes, and those loans were not forgiven.
00:09:27To build on that success, we need more innovation in government.
00:09:32In 2022, we recommended that one way to do that.
00:09:36Congress could establish a permanent analytics center of excellence to unlock tools that more efficiently identify and prevent fraud and improper payments.
00:09:45My final point is that harnessing innovation also requires a federal workforce that has the right skills.
00:09:52But agencies continue to face barriers in hiring, managing, and retaining staff with advanced technical skills.
00:09:59This is another area where GAO has made recommendations, and we have explored innovative big-picture ideas like establishing a digital services academy.
00:10:11Rapid advances in AI capabilities hold great promise.
00:10:15GAO believes that the federal government can and must simultaneously realize those opportunities and do so upon a firm foundation of reliable data in a digitally skilled workforce.
00:10:27Chairman Schweikert and Ranking Member Hassan and members of the committee, thank you, and I'd be happy to respond to any questions.
00:10:33Thank you, Dr. Thomas.
00:10:34Dr. Miller.
00:10:35Chairman Schweikert, Vice Chairman Schmidt, and distinguished members of the committee, thank you for allowing me to join you and share my views on how AI and technology
00:10:45can improve government efficiency and reduce broad waste and abuse.
00:10:50I'm going to focus on operational solutions targeting the Medicare and Medicaid programs, recognizing that these comprise over $1.5 trillion in annual spending.
00:11:02I should note that today I'm here in my personal capacity and the views expressed are of my own and don't necessarily reflect those of Johns Hopkins University,
00:11:10the American Enterprise Institute, the Medicare Payment Advisory Commission, or the North Carolina State Health Plan.
00:11:17That aside, ethos of DOGE, asking questions, reengineering processes, these are things that we need to do definitely in government health programs.
00:11:27So there are three areas that I think that we should focus on.
00:11:30The first is Medicaid improper payments.
00:11:32In 2020, we had the global pandemic, tens of thousands of people were dying every day, and Congress implemented continuous Medicaid coverage,
00:11:40thinking that, you know, if you're going to potentially have COVID and need hospitalization, coverage shouldn't be the issue.
00:11:47In 2023, states started redetermination for 20 million beneficiaries.
00:11:53CMS noted in 2022 that 76% of improper payments were due to eligibility.
00:11:59That's $61 billion a year.
00:12:01Now, I looked this up and Medicaid eligibility requirements are codified in statute with little discretion.
00:12:08So if you think about it, redetermination, and then on the other side, initial eligibility assessments are two sides of the same process.
00:12:16We should actually look to automate benefits eligibility.
00:12:20We might have to have human review first, followed by an auditing period, and then a move towards fully autonomous review.
00:12:27With the simple principle being, if you're entitled to Medicaid benefits, you should be able to get them.
00:12:32And if you are not entitled to Medicaid benefits, you should not be on Medicaid.
00:12:37The second area I'd focus on is Medicare Advantage diagnosis coding, which I think we've all seen the Wall Street Journal articles and the large plethora of news about this marketplace.
00:12:48So if you think about it, in fee-for-service Medicare, it's like running a pizza shop.
00:12:52You're paid if you make more pizzas.
00:12:53If you make a slightly fancier pizza, you're also paid a little bit more.
00:12:57In Medicare Advantage, it's a bit different.
00:12:59You get risk-adjusted capitated payment or a population-based payment with the risk adjustment from diagnosis coding.
00:13:06That way, if plans take care of sick people, they get paid more.
00:13:10If they take care of healthier people, they get paid less.
00:13:12It's no surprise that across these two programs that coding intensity is different.
00:13:17Coding intensity has three components.
00:13:19One is fraudulent coding.
00:13:21The DOJ picks you up in a black car and takes you to jail.
00:13:25Two is up-coding, subject to debate but likely known.
00:13:29The third is sort of clinically appropriate coding intensity.
00:13:32Better communication about diagnostic specificity or severity.
00:13:37So we've sort of studied this problem for two decades.
00:13:40You can pick your favorite health policy analyst, economist, actuary, committee or commission, and you can get different numbers.
00:13:47No one has measured those three components.
00:13:49The research has shown that it varies across plans.
00:13:52I would argue that if we have predictive typing and autocorrect on our iPhones, we can automate diagnosis coding across the Medicare program.
00:14:02Do we want doctors spending time with patients or do we want them doing diagnosis coding for whether you have CKD, chronic kidney disease, stage 3A or 3B?
00:14:10I think we can all agree that we want doctors taking care of patients.
00:14:15And so if we automate that, we will eliminate coding disparities across both programs, increase trust, eliminate the need for chart reviews and health risk assessments,
00:14:26and eliminate the wastage of both physician times and government payments.
00:14:31So I would say that for Medicare Advantage, we should focus on solving problems for Americans instead of measuring them.
00:14:37And then the last area I want to talk about is prior authorizations.
00:14:40We spend an estimated $265 billion in the healthcare industry on administrative costs.
00:14:46Depending upon who you ask, prior authorization can be an appropriate tool to steer care,
00:14:52or it can be an inappropriate tool to delay care, increase costs, decrease access.
00:14:57And you'll have sit around, you'll have stakeholders on either side showing up telling you one thing or the other.
00:15:03We've also talked about this for probably longer than I have been alive.
00:15:08So I would say that we should start with process reform, fix the process.
00:15:13So let's put the fax machine lobby out of their misery.
00:15:17Same for, you know, PDF submission, et cetera.
00:15:20We should use CMS and ONC to drive automation of data submission and have it be built directly into electronic health records.
00:15:30So I think we have a lot of opportunity to practically make government more efficient, reduce waste,
00:15:35and also make the delivery system better.
00:15:38And thank you for the time.
00:15:40Chairman Schweikert, distinguished members of the Joint Economic Committee.
00:15:47Thanks for the opportunity to be here to discuss how artificial intelligence can transform government efficiency.
00:15:54America leads the world in advanced AI technologies.
00:15:58Private companies are rapidly deploying these tools to boost productivity and eliminate waste.
00:16:03But our government is only beginning to tap their potential.
00:16:07Of the 2,100 AI use cases identified in January of this year across 41 U.S. federal agencies, I could only identify 17 operational systems that directly target fraud, waste, and abuse.
00:16:21And most of these are on a small scale.
00:16:24This represents both a challenge and an extraordinary opportunity.
00:16:29The Trump administration has recognized this.
00:16:31Just last week, the Office of Management and Budget issued two detailed memos that highlight these benefits and set forth a path both for federal government use and the procurement of AI technologies.
00:16:46Consider the staggering costs of fraud.
00:16:48We've already heard some of these numbers.
00:16:51The federal agencies reported $162 billion in improper payments last year alone across 68 programs, with Medicare and Medicaid accounting for $85 billion of that.
00:17:01And GAO has estimated in the past that an annual fraud loss of potentially as high as $521 billion.
00:17:10That's potentially half a trillion dollars of taxpayer money each year that never reach its intended purpose.
00:17:17AI technologies offer some powerful solutions.
00:17:19We hear a lot about large language models and chatbots, but I'll talk about a few specific tools that I think have great application here and then two specific areas in which, two broad areas of government in which they have some applicability.
00:17:34Again, the point here is that these are well deployed in the private sector and could be used in the government.
00:17:40So anomaly detection is a technology that allows you to quickly spot unusual activities by learning what constitute normal behavior and then identifying abnormal behavior.
00:17:52Financial institutions like HSBC have found that this approach identifies two to four times more financial crime with 60% fewer false positives.
00:18:01Natural language processing, of which LLMs are one type of application, can digest mountains of text documents, emails, and support tickets far faster than humans.
00:18:12This technology can streamline compliance, enhance customer service, and can reveal fraud that's hidden in mass amounts of text.
00:18:19Robotic process automation uses software bots to handle routine tasks.
00:18:25It often is used to connect legacy systems, of which obviously government institutions have many.
00:18:31And businesses that implement this see an average efficiency gain of up to 30% while reducing processing errors by 20 to 40%.
00:18:39And then finally, the last category of technology, although there's many others.
00:18:43Graph analytics map relationships that entities have across a bunch of different systems.
00:18:49And so insurers right now use graph analytics to identify fraud rings and coordinated abuse that you could not find if you were only looking at a single system at a time.
00:18:59So these technologies offer tremendous potential across government.
00:19:03But let me give you two critical areas where I think they could offer some real specific benefits.
00:19:08So in benefits administration, obviously I mentioned Medicare and Medicaid, but many other areas of benefits.
00:19:16AI can swiftly identify fraudulent claims while speeding service to legitimate beneficiaries.
00:19:21They can verify documentation, automate eligibility verification, and help agencies forecast their needs to optimize resource allocation.
00:19:30And then in procurement and contracting, these tools can flag high-risk vendors and contracts before the money goes out the door.
00:19:40They can detect vendor collusion, and they can automate repetitive procurement tasks, streamlining these functions both for the contractors and the government agencies that rely on them.
00:19:52Now, the path forward here is going to require updating some legacy systems, safeguarding data privacy, and addressing some important policy considerations.
00:20:01But the potential benefits are profound.
00:20:04And that's why when I look at AI's capabilities in this space, I don't see just a futuristic vision.
00:20:09I see there's practical, proven technologies that are already transforming private industry that could immediately strengthen government operations.
00:20:18Implementing these solutions means ensuring public funds reach their intended purposes.
00:20:23It means faster, more transparent services for citizens.
00:20:27And it means substantial savings that could be redirected to higher priority needs or returned to taxpayers.
00:20:34The question before us is not whether AI can improve government efficiency.
00:20:38The evidence from the private sector is overwhelming that it can.
00:20:41The question is whether we have the vision and the commitment to embrace these innovations for the benefit of all Americans.
00:20:47Thank you, and I look forward to your questions.
00:20:52Good afternoon, Chairman Schweiker, Ranking Member Hassan, and members of the committee.
00:20:56Thank you for inviting me to discuss how technology can improve government efficiency and reduce waste, fraud, and abuse.
00:21:02I serve as the Executive Director for the Council of the Inspectors General on Integrity and Efficiency, or CIGI, the membership organization for more than 70 federal IGs.
00:21:12The IG community has a nearly 50-year history of conducting effective government oversight and detecting and preventing waste, fraud, and abuse.
00:21:20The Inspector General Act of 1978 established IGs and empowered them to provide independent, objective, and impartial oversight of agency programs and operations.
00:21:30IGs have a dual reporting responsibility, reporting to both their agency head and to Congress.
00:21:35We have a long-standing relationship with Congress and a shared goal of ensuring that taxpayer dollars are spent wisely and effectively and that government benefits are administered correctly.
00:21:46IG findings are based on facts and applicable law and guided by OIG standards.
00:21:51Trained auditors, investigators, inspectors, evaluators, IT specialists, and attorneys work every day to promote efficient and effective government operations.
00:22:01In fiscal year 2024, the IG community collectively totaled over $71 billion in audit and investigative accomplishments, generating a return of about $18 for every $1 invested in OIGs.
00:22:14For the government to see the full value of this oversight, agencies must take action to resolve OIG recommendations.
00:22:21There are currently over 14,000 unimplemented OIG recommendations, many of them related to improper payments.
00:22:28If implemented, such actions would improve payment integrity across dozens of federal programs.
00:22:34For example, the Department of Labor OIG identified over $45 billion in potentially fraudulent pandemic unemployment benefits, and the OIG has engaged the department in resolving these high-risk payments.
00:22:46However, there are still outstanding recommendations in this area.
00:22:49While all fraud is an improper payment, not all improper payments are fraudulent.
00:22:54Other common causes include administrative mistakes, recipient self-reporting issues, and policy limitations.
00:23:01For instance, the Social Security Administration OIG estimated over $7 billion in supplemental security income errors because SSA did not follow policies or use all available tools to prevent, detect, and recover overpayments.
00:23:15SSA OIG has made several recommendations to reduce and collect the overpayments, but the recommendations have not yet been resolved.
00:23:23The use of technology is critical to the IG community's oversight work, and IGs have significant experience leveraging data analytics and artificial intelligence.
00:23:32While these tools do not replace investigative or audit work, they enhance IG's ability to identify anomalies for further review.
00:23:40Last year, the Treasury Inspector General for Tax Administration leveraged data analytics to help the IRS prevent $3.5 billion in potentially improper employee retention credits and sick and family leave credits.
00:23:54It takes a notified IRS officials of the scheme, who then put controls in place to guard against similar fraudulent claims.
00:24:00The Postal Service OIG uses data-driven insights to analyze millions of customer complaints to recognize trends and identify potential mail delivery or theft issues.
00:24:11And the Office of Personnel Management OIG is using machine learning to inform which health plans the OIG should audit in the Federal Employee Health Benefits Program.
00:24:21To continue our effective oversight efforts and improve payment integrity, sustained access to agency data and analytics resources are critical.
00:24:30The IG community's top legislative priority is the establishment of a permanent, scalable data analytics platform to aid OIGs in detecting and preventing fraud and improper payments.
00:24:40In 2021, Congress provided CIGI's Pandemic Response Accountability Committee, or the PRAC, with the authority and funding to create a data analytics center.
00:24:50The result was the Pandemic Analytics Center for Excellence, or the PACE.
00:24:55The PACE has proven to be successful in identifying improper payments and fraud in pandemic programs.
00:25:01In one recent case, the PACE helped an OIG develop risk models to review some 2,400 pension plans, which led to the recovery of over $165 million.
00:25:11This recovery for taxpayers far exceeds the $120 million that Congress appropriated for the PRAC to operate for the last five years.
00:25:20However, with the PRAC scheduled to sunset on September 30th, 2025, the IG community urges Congress to sustain the PRAC and the PACE's data analytics capabilities.
00:25:30Last month, the House Committee on Oversight and Government Reform advanced legislation to sustain the PRAC and its capabilities, which we support.
00:25:39We look forward to working with Congress on this important reform to ensure the IG community is equipped with advanced technology and data analytics tools.
00:25:47In conclusion, the federal IG community remains committed to conducting effective program oversight, making recommendations that can improve government efficiency,
00:25:55and leveraging analytics and AI-supported tools in its efforts.
00:25:59Thank you again for inviting me to testify, and I'm happy to answer any questions.
00:26:02Thank you for all of you.
00:26:04Because I have hours of questions, I'm going to go at the very end.
00:26:07Senator Schmidt, you get to start with questions.
00:26:12Thank you all for being here.
00:26:17Thank you all for being here.
00:26:18Thank you, Mr. Chairman, for holding this hearing.
00:26:21I think some of the recent findings from DOGE, I think, has put this even more up front in the public consciousness.
00:26:30And there's a lot of great work that's happened before.
00:26:34And I'm not going to go through.
00:26:36I was going to talk about, you know, the money that's sent overseas for Guatemalan sex changes and DEI in Burma, but I've done floor speeches about that.
00:26:46I really want to use the time that I have with asking some questions that I think, I can't imagine there's a ton of disagreement as far as trying to understand how we're at where we're at.
00:26:58So, Dr. Miller, you brought this up in the healthcare space.
00:27:05I think one of the great frustrations for policymakers is that it just, the enormity of some of the problems that exist are so huge that sometimes I think that inhibits people's ability to tackle them effectively.
00:27:18What do you think, so you've got, you've had some great recommendations about some things that could happen and whether it's coding or finding fraud on the front end or the back end in a much more meaningful way.
00:27:31It feels like, as you said, you've been having this discussion longer than you've been alive.
00:27:36I can relate to that.
00:27:38What is missing?
00:27:40Like, what, we know what the problem is.
00:27:43Recommendations are made, and I want to ask you this question too.
00:27:47What is, what's the disconnect?
00:27:49Like, how, how are the controls not actually put in place?
00:27:53How do, how does this keep happening?
00:27:55Why isn't, you know, and Doge, anyway, I'm going to stop talking and let you answer the question.
00:28:00So, two reasons.
00:28:01One is, whether it's fraud or an improper payment, that's someone's revenue and that person who has revenue doesn't like it when you take their revenue away.
00:28:10So that's a political issue.
00:28:12After that, there's the operational issue.
00:28:16Which is often where measure the problem, we issue a nice report, write a nice paper, and then we don't get the agency wheels going to address it.
00:28:26So we need to have oversight, public attention, and energy directed at getting the agencies to execute an operational plan.
00:28:37If they don't have an operational plan, have an operational plan.
00:28:40For example, we know that there's often fraud and procedures, and if a procedure takes two hours and you bill 20 of them in a day, that's clearly not possible.
00:28:50We should have prepayment claims editing in Medicare and Medicaid to prevent that from happening.
00:28:57We don't.
00:28:58So I would say aggressive and assertive oversight over agencies for what is the, what are the operational steps that they're going to do, and then continued oversight to make sure that they execute on those steps.
00:29:10So are AI, because it's subject to the hearing, are AI tools being used currently, outside of DOGE, right, because they're looking under the hood.
00:29:20Are those sort of tools being used effectively currently in these agencies?
00:29:27I would say no.
00:29:28Right.
00:29:29That's, I think, the right answer.
00:29:30I wanted to ask you, so you've also talked about instances where recommendations were made.
00:29:36In some, like the PACE sort of oversight, you talked about how that was effective.
00:29:41Maybe that's the model.
00:29:42I don't know exactly, but what is the disconnect?
00:29:46Is it that Congress isn't calling these people in enough?
00:29:50Is it that if it's not, you know, there's not, there's not a point person within the agencies that are actually responsible for this?
00:29:58How would you diagnose the problem to be able to fix?
00:30:01I think Dr. Miller said it well that there's a, there's a plethora of information out there.
00:30:06There's recommendations that agencies can implement, and where Congress can help is increasing transparency and visibility of those recommendations,
00:30:14and talking directly to the agency.
00:30:16You know, what's prohibiting you from implementing it?
00:30:18Is it a resource issue?
00:30:20Is it a time issue?
00:30:21Is it a need to continue customer service, and it's just trying to implement these implementations that's holding us back?
00:30:27I think the information is there.
00:30:28It's understanding what's holding the agency back.
00:30:31Are there any other, I mean, obviously there's the power of the purse, and there's the appropriations process.
00:30:36Are there any other, as you guys have found, any of you, are there any other hammers that are out there that would actually hold some of these folks their feet to the fire to actually implement these things?
00:30:47I might flag something quickly from the OMB guidance that just came out last week.
00:30:53One of the things that they recommend on AI implementation within agencies is that the responsibility for addressing and owning the risk be on the person who is responsible for the mission of the process that AI is being deployed for.
00:31:09And so that type of personal accountability where there has to be somebody actually signing off and owning the decision, I think is the sort of thing that keeps new technologies that are coming in under the right amount of scrutiny and analysis.
00:31:23And so I don't know that that, how you retrofit that, but for new technologies, I think that's a good approach.
00:31:29Thank you, Mr. Chairman.
00:31:31Well, thank you, Mr. Chair.
00:31:34And again, thank you all for your testimony.
00:31:36I caught most of Mr. Canarsa's and I'm sorry I missed yours, but I want to start with a few questions.
00:31:42Mr. Canarsa, inspectors general play a key role in identifying and rooting out fraud across the federal government.
00:31:48As part of their efforts, they often encounter barriers when they try to work together across agencies and share data that could prevent fraud.
00:31:55For instance, an inspector general at one agency might not have the staff or time needed to access data on fraud that's stored at multiple other agencies.
00:32:05I know that your organization wants to establish some centralized data analytics capacity that the investigators could use to further detect and prevent fraud across the government.
00:32:15So this is a two part question.
00:32:17One, what are some of the most promising examples of how inspectors general are already sharing data to prevent fraud, such as by working together to stop criminals who have multiple shell companies that have been defrauding our small business loan programs?
00:32:32So that just, why don't we start with that?
00:32:35Sure.
00:32:36Yeah.
00:32:37Sure.
00:32:38Thank you for the question.
00:32:39And Senator Hassan, I wanted to thank you for working with Senator Joni Ernst on the IG Senate caucus.
00:32:44I appreciate your efforts and look forward to working with you both on that.
00:32:48To the question, several examples going back to the start of the pandemic and want to give credit to the Small Business Administration OIG and Department of Labor OIG,
00:32:58who had oversight over the PPP and the EIDL programs and then the Pandemic Unemployment Insurance Program, which obviously at that time, it was a need for those funds to go out at the time.
00:33:10Right.
00:33:11But it left it right for fraud.
00:33:13And where the, where the prac came in and helped was to identify those, a lot of those suspicious claims that were clustered and which led to, in those cases, groups that were putting in multiple claims, you know, because they saw some go through.
00:33:28So they kept going for them.
00:33:29Right.
00:33:30In both instances, both SBA OIG and DOLOIG have successful fraud investigations to the tune of millions of dollars that were, that were found and prosecutions continued from there.
00:33:44Great.
00:33:45And what can Congress do to make sure that that sort of cooperation gets easier while also protecting people's privacy?
00:33:53I think, as I mentioned, the expanding and enhancing the potential of the prac's pace.
00:34:00You know, if we can do what they've shown in just the pandemic realm programs, if we can do that across federal programs, make it easier for, as you mentioned, for agencies and their OIGs to match in a central location.
00:34:12Again, with, with all privacy protections and considerations in place, that just makes it easier to go to one stop instead of each agency trying to make these agreements one off.
00:34:23Right.
00:34:24And in my experience as a governor, it, you know, one of the hard things is finding actual staff and resources to focus on this so that if we can share in the, in the effort makes a big difference.
00:34:34Dr. Thomas, your position as the chief scientist of the government accountability office gives you a good vantage point to see how the federal government deploys technology, especially when it comes to strengthening government, government effectiveness and saving taxpayer dollars.
00:34:49What are areas of the government that have the highest potential for reducing waste, fraud and abuse through the deployment of cutting edge technology, including AI?
00:34:59Yeah.
00:35:00So if you look at our, our government wide fraud report that we just put out, we talk about the, the different programs that in the, there's a range that goes with it.
00:35:09So I think as earlier mentioned, half a trillion dollars.
00:35:11That's the high range.
00:35:12Those are for high risk programs.
00:35:13You know, think about like COVID style programs.
00:35:16The lower range is really analyzing the lower risk programs.
00:35:20We did a modeling exercise using Monte Carlo to understand what that range looks like.
00:35:25Essentially what it comes down to additional transparency, obviously within the payment systems and making sure that the agencies are using the analytics and sharing the use of that analytics, as well as what has already been talked about earlier that we support is setting up a center of excellence that supports the fraud world.
00:35:44So that the, across the government, all of these agencies can work together, can share data that matters and share tools that identify this flags of potential fraud.
00:35:54The important part though, that I want to make sure we don't miss is that the analysts play a critical role.
00:35:59Yeah.
00:36:00The analytics only identifies a flag that some may, something might be associated with fraud.
00:36:04It is the analysts themselves that actually goes and does the investigation.
00:36:07Right.
00:36:08That's very helpful.
00:36:09I have lots more questions too, but why don't I yield back and then we can keep going.
00:36:13Yeah.
00:36:14Thank you, Mr. Chairman.
00:36:15I want to thank you all for being here today.
00:36:18I wanted to point out that in my district, we're experiencing a real problem with treasury checks being stolen.
00:36:26And obviously this is not something that's unique to my district, although it's been heavily impacted.
00:36:31As you've mentioned, the GAO estimates a total of 162 billion in improper payments in fiscal year 2025 and more than 2.8 trillion since 2003.
00:36:42So you have the issue of improper payments and then you have those of the fraudsters and the folks that are actually physically stealing these checks, washing them, reissuing them, and cashing them.
00:36:56And in my district alone, we've had roughly 400 constituent cases of stolen federal checks that total $5.4 million.
00:37:06So this is a big issue, right?
00:37:08If we multiply that by all the members and the senators in this body, you're talking about a substantial amount of money.
00:37:15And I would love to hear from any of you on what you think could be a viable solution.
00:37:21Obviously, moving people to direct deposit is one thing.
00:37:24We're looking at.
00:37:25The president did issue an executive order.
00:37:26We recently passed legislation out of the House to require IRS to reissue payments if the consumer wants or the taxpayer wants via direct deposit.
00:37:37Maybe tracking checks so we know what point these checks are being stolen.
00:37:44And then AI, as you're discussing today, can come in handy as well.
00:37:47So we'd love to hear from any of you on any ideas or solutions you may have or what you've seen work potentially in the private sector as it relates to financial crimes that could be applied here.
00:38:02Yeah, so I don't have an answer for the direct question of what do you do about washing checks, Congresswoman, but I appreciate the question.
00:38:11You know, within how do you, you know, so we're talking about really fraud, because fraud is where you get into the point of someone's purposely trying to defraud the government, which is what this is about.
00:38:23You know, analytics certainly and AI could certainly help this by tracking the actual payment beyond just the check, making sure that the individual is actually receiving it,
00:38:31providing some sort of methodology to check back.
00:38:34Those are types of things that the private sector uses that can then be repurposed potentially for the government.
00:38:39It is not an area that we have looked into though at GAO.
00:38:41Mr. Chilson?
00:38:43So obviously taking the middleman, I mean, obviously taking the middleman out, which is the mailbox, I guess, and doing direct deposit makes a ton of sense.
00:38:54It's not really an AI approach.
00:38:59It is obviously an established technology, but there you're relying on the commercial sector, which does have a lot of capabilities to identify, you know, when electronic payments are working and when they're not.
00:39:13Electronic payments are not fraud free, but the companies have spent a lot of money to identify how to make sure that money moves properly between accounts.
00:39:21And so I do think that the more you can move people to that type of approach, the better it will be.
00:39:26Dr. Thomas, in your testimony, you highlighted the need to extend the Social Security Administration's authority to share its full death data master file with the Department of Treasury, which is set to expire next year.
00:39:39And I would like to know your thoughts on how much this has recovered, how successful it's been.
00:39:45It's my understanding it's about $31 million in payments.
00:39:48And can you highlight any other multi-agency data sharing that needs to take place?
00:39:54I would love to, but I want to make sure I give you exact information.
00:39:57So if you don't mind, I'll get back to you on that.
00:39:59Okay, no problem.
00:40:00And then one more question for you.
00:40:01What is the most significant barrier stopping the federal government from implementing these advanced AI technologies to eliminate waste, fraud and abuse?
00:40:09Yeah, so I think there's two areas that I want to particularly highlight.
00:40:13And, you know, this is one of the biggest questions out there.
00:40:16And, you know, the first one I think we've already talked about quite a bit, which is just the partitioning of data,
00:40:21trying to make sure that data is available to the anti-fraud community so that we can track people across the different parts of the government
00:40:28and track the information so that it can actually be used for large types of analytic models.
00:40:34The second is actually just as important and probably needs to happen before the first,
00:40:37which is a foundational data that is validated.
00:40:40It shows these are actual indicators of fraud or improper payments.
00:40:44We can't get into a habit of marking something that is not fraud and making a guess at it.
00:40:49We do actually need to go through the entire process with the IGs to make sure that something has actually been validated
00:40:54and put that into kind of a gold standard data set.
00:40:58That is going to produce the best outcome for our analytical tools.
00:41:06Mr. Chairman, thank you very much.
00:41:07And Senator Hessen, thanks for holding us together.
00:41:11And thank you all very much for being here.
00:41:13I really, really appreciate this.
00:41:15So many questions I want to ask.
00:41:22But let me just start with, you know, we've watched Elon Musk and his Doge team recently go after waste, fraud, and abuse.
00:41:30And you've laid out the case very strongly for why that is necessary.
00:41:34You know, these many hundreds of millions of billions of dollars out there.
00:41:38Let me just point out, though, that most of this seems to be improper payments or folks stealing checks.
00:41:44I'm not quite sure the connection between the huge amount of waste, fraud, and abuse you've identified
00:41:51and firing half the people at the Department of Education, laying off 20,000 people at HHS,
00:41:56getting rid of the U.S. Institute of Peace, firing all probationary employees.
00:42:00None of that seems to address the waste, fraud, and abuse that you are trying to find very constructive, honest ways to deal with.
00:42:07And I still remember fondly, I just figured out it was 50 years ago that the little family business
00:42:15where we were doing all of our records by hand with a pencil and an end key, 10 key adder.
00:42:19And on December 1st, 1975, I moved it to the internet.
00:42:24It was a keyboard the size of an electric thing online, so it took five minutes between entries when you pushed the key.
00:42:31However, it made possible a degree of productivity increase that's just unprecedented, including much, much less waste, fraud, and abuse.
00:42:41So, Chairman Schweiker and I share the vision of what this could do to federal government and to Medicare and Medicaid to be able to bring that technology here.
00:42:51In the meantime, I'm very concerned about DOGE, which I want to succeed, but the use of unapproved technologies without sufficient oversight,
00:43:00the use in high-impact areas like federal employees' employment status, and the lack of basic security for non-government datasets,
00:43:08instead profiteering off of sharing data with commercial models.
00:43:11I prepared a letter with many of my Democratic colleagues, which you're going to send to OMB.
00:43:15And Mr. Chair, I'd love to submit that letter to the record.
00:43:19And then ask Mr. Chilson, in your testimony, you talked briefly about things like audit standards, up-print training and investment.
00:43:27Which existing laws and specific new policies would you promote to encourage the responsible adoption of AI?
00:43:36So, I'm most familiar with regulation in the commercial sector, having spent time at the Federal Trade Commission.
00:43:43But I am familiar with, at the FTC, some of the practices of every, you know, privacy officer that is within agencies.
00:43:51And I think the recent OMB memos, again, if you haven't seen those, I think they offer a lot of the safeguards that should be undertaken as an agency adopts technology into use in pursuit of its mission.
00:44:08These memorandum replace some Biden-era memorandum, but they cover a lot of the same area, maybe with a slightly different focus.
00:44:19One that more encourages the adoption of this technology in pursuit of the mission, but offers the same sorts of guardrails in some cases that will help make sure that this protects privacy,
00:44:32that it pursues the mission of the agency, and that it's effective.
00:44:35And so, some of those principles in there, I think, are a good start, although they're, you know, they're guidance, not regulation.
00:44:41Yeah, thank you, Mr. Chilson.
00:44:43And Dr. Miller, one of my favorite agencies in Congress is AHRQ, the Agency for Healthcare Research and Quality.
00:44:49Because their mission is all about how to deliver healthcare more effectively and not spending money on costly, wasteful treatments.
00:44:56Someone once told me that if you want to look at efficiency in federal government, AHRQ is the poster child.
00:45:01And it's one of the key agencies that looks at over the $4.5 trillion spent annually on healthcare.
00:45:07And it's even more important when you think about artificial intelligence supplying it.
00:45:11But, again, guess what?
00:45:13Much of AHRQ was gutted by the reduction in force at HHS last week.
00:45:17And the people that we need to do HI for healthcare were actually fired.
00:45:23The RIF apparently eliminated most of the staff in the office of the Chief Information Officer.
00:45:28And then they attempted to relocate most of the executive staff to, like, the Indian Health Service.
00:45:34I'm sure they thought they could outsource the IT services, but you need people on the ground that understand tech policy and cybersecurity.
00:45:43What would you do to make sure that we have the folks in the agencies that can make sure that we have the expertise to advance AI adoption?
00:45:53I think AHRQ is more of a research and a research contract organization as opposed to a regulations or operations organization.
00:46:02So I would say that CMS, which is functioning as the payer and the regulator of the private payers executing the public program, needs the digital health, AI, machine learning expertise built into CMS, into each of the centers, whether it's the Center for Medicaid, whether it's the Center for Medicaid and CHIP services.
00:46:21I don't think that that is AHRQ's primary function historically.
00:46:25Okay.
00:46:26Well, great.
00:46:27Well, my time's up, but thank you very much.
00:46:28Thank you, Mr. Biden.
00:46:29Thank you, Mr. Chairman, for holding this hearing and for your work on this committee.
00:46:37Today, I'm afraid that rooting out waste, fraud, and abuse has become divisive here in Washington, but it's exactly what the Americans want us to address.
00:46:47Before this year, rooting out waste, fraud, and abuse was actually a bipartisan endeavor and talked about and had broad support.
00:46:56It appears that now there's a number of lawmakers, bureaucrats, and media that actively support the status quo, which in essence is on the side of allowing the waste and fraud to continue.
00:47:06We've had decades of inspector general reports that have outlined the reckless misuse of government funds with little or no appetite to address that.
00:47:15And you mentioned that earlier today in terms of that there was the status quo or the push on the status quo was to resist the loss of recipients of the payments that were making through that.
00:47:31The new administration is shattering some of the status quo and tackling the problem with their directive that got from the American people last November.
00:47:41As the chair of the Social Security Subcommittee here in the House, I want to just highlight that for everybody in America that every dollar that's saved by eliminating waste, fraud, and abuse is a dollar that can go towards benefits to seniors that have earned them or for benefits in other programs as well.
00:48:00Typically, things have been slow here in Washington to change with just the bureaucracy and the size of the government.
00:48:09If you contrast that with the private sector, U.S. businesses, if they don't innovate and adapt, they die.
00:48:15In D.C., as Ronald Reagan once quote, a government bureau is the nearest thing to eternal life we'll ever see on earth.
00:48:23Some of the advancements that we're seeing with artificial intelligence are already disrupting industries throughout the country,
00:48:29whether it's in the medical field, manufacturing or technology.
00:48:32The federal government needs to be looking at ways that we can actually be more responsible and use AI.
00:48:38Dr. Thomas, based on data from 2018 to 2022, GAO estimates that aggregate fraud was between $233 billion to $521 billion each year.
00:48:51On the low end, that means that the government's lost more than a trillion dollars in a five-year period and maybe as much as $2.6 trillion.
00:49:00It's unfathomable that people would be against finding and eliminating this blatant theft of taxpayer dollars.
00:49:09So what, from a GAO's perspective, are you doing to implement machine learning and data analytics to help reduce that fraud
00:49:16and how other federal agencies can work to get that number closer to zero?
00:49:22Yeah. So within our data analytics programs, we are developing tools for our analysts to identify the flags that I mentioned earlier
00:49:30of what are potential indicators of fraudulent events, fraudulent programs, fraudulent payments, and making those tools potentially available to the rest of the agencies to use.
00:49:40And we continue to work with the agencies to help them adopt the recommendations that we put forth in each of our reports.
00:49:46When we talk about fraud across the government as an estimate, that comes from information that we receive from each of the agencies.
00:49:53And as we work with that, we develop a whole bunch of recommendations and we follow up with them and, you know, adopting those recommendations,
00:50:01we believe will bring down the risk of fraud.
00:50:04Additionally, we want them to use our fraud risk framework that we have implemented and shared with them.
00:50:09It allows them to look at programs and identify where potential fraud could exist if that program was implemented in that particular way
00:50:18and make adjustments to try to, just as you talked about, stop fraud from happening before it happens.
00:50:23Thank you. Mr. Chilson, you know, I think when we talk about rooting out waste, fraud, and abuse,
00:50:29sometimes a false narrative gets promoted that the federal government is taking away benefits from Americans that have rightfully deserved those.
00:50:38And trying to alleviate some of this fear mongering, can you help dispel some of those fears and explain how AI can more effectively target those bad actors,
00:50:47target the improper payments to ensure that our government programs provide the resources to the right beneficiaries?
00:50:53Well, accuracy is really important. And what we've seen in the private sector is that these tools can help both eliminate fraud,
00:51:02but also serve the people who are rightfully getting benefits more quickly and accurately.
00:51:10And so I do think that we have opportunities in this space. The financial sector does this all the time.
00:51:15And what it's found is that, you know, you may remember you got lots of calls, maybe five years ago,
00:51:22you got lots of calls anytime you like went on a trip with your credit card.
00:51:25Well, now they can identify when there's fraud with a lot fewer of those false positives.
00:51:30And that's a convenience to the people who are getting benefits as well as those who are, you know, as well as addressing the people who are fraud.
00:51:37Well, thank you. And I've run out of time like so many others.
00:51:40We have so many questions that could ask each of you and yield back, Mr. Chairman.
00:51:46Thank you, Chair Schweikert and Ranking Member Hassan.
00:51:49Really privileged to be on this committee and to be part of today's hearing on this very important topic.
00:51:55Obviously, when we talk about waste, fraud and abuse, we all want to reduce this.
00:51:59We all want to eliminate as much as possible.
00:52:01Ensuring the taxpayer dollars are spent effectively is part of our mission here in Congress.
00:52:07And so very excited to hear about the possibility of what we can do to try to rein in some of these waste, wasteful actions.
00:52:16Every year, the federal government disperses trillions of dollars, as we know, across thousands of different programs,
00:52:22health care, infrastructure, defense, disaster relief, and on and on and on.
00:52:26And of course, with this level of spending comes the inevitable challenge of effectively managing waste, fraud and abuse.
00:52:32Historically, large organizations, including, I just want to note, private companies, big corporations,
00:52:38have all had to deal with this problem of waste, fraud and abuse.
00:52:41And they've done so through manual audits and control systems.
00:52:45But the rapid development of technology over the past few years, including AI, as we're talking about today,
00:52:50give us a really transformative opportunity to try to run government better.
00:52:54We're talking about some very, very exciting potential here.
00:52:57That being said, I think it's important to also recognize that technological advances like AI are only going to be as good as the parameters we set for them.
00:53:06And these are fundamentally normative parameters we're talking about.
00:53:09So when we think about efficiency, we need to talk about the goals and timelines that are our mission here.
00:53:15Are we talking about just reducing costs?
00:53:17Are there other goals involved?
00:53:19What's the timeline we're looking at for efficiency or cost cutting?
00:53:22Is it immediate?
00:53:23Is it a year out?
00:53:24Is it 10 years out?
00:53:25All of these things matter.
00:53:26And I just want to give you an example here.
00:53:27When we talk about Social Security, despite a number of really bad and false and debunked claims about Social Security fraud recently,
00:53:36we know that Social Security has been a remarkably effective program and an efficient program.
00:53:40A very, very low percentage of fraud, always under 1% under any analysis.
00:53:45A very low percentage of overhead spent on that.
00:53:48Less than 0.5% of the total program is spent on personnel and overhead.
00:53:53And until this year, the Social Security Administration was renowned for its remarkable ability to deliver trillions of dollars of checks consistently, securely, and on time.
00:54:04Are there ways to cut back on Social Security?
00:54:07Yes, of course.
00:54:08And we've seen Elon Musk and Doge do just that, attempting to fire 12% of the Social Security workforce.
00:54:14This has, though, led to lots of reports of delayed and missing checks, difficulty for seniors, including ones that have called my office,
00:54:21trying to find personnel that will deal with their problems with Social Security.
00:54:25The question, of course, is what do we mean by efficiency when it comes to Social Security?
00:54:30Are we talking just about cutting costs?
00:54:32Are we talking about cutting costs now?
00:54:33Are we talking about other goals here?
00:54:35I would submit to you that I think what Social Security has done is get checks out to people.
00:54:40That is its mission.
00:54:41And so, again, when we think about AI and how it can prove efficiency across different agencies, those goals, the timeframe that we're talking about, are important.
00:54:51I want to give you another example.
00:54:53I represent a vibrant life sciences industry, and I've heard a lot of growing concerns about the major cuts that have been made to staffing at the Food and Drug Administration,
00:55:01and that these will lead to longer approval times for new drugs and devices.
00:55:05And, in fact, if you look at the FDA structure, these staff are actually paid for largely by user fees from the companies seeking approval.
00:55:13So, actually, the cost, other than if you're looking at the very immediate timeframe, is zero.
00:55:18But if your priority is to look for immediate spending cuts, the administration, the terminations of lots of FDA staff might make some sense.
00:55:26But beyond just the immediate present, it doesn't make any sense.
00:55:29And if you're looking at any other goals, including FDA approval, rapid approval of drugs and devices, these cuts look very inefficient.
00:55:36So, I guess my first question to you all, just maybe I'll start with Dr. Thomas.
00:55:41When we talk about efficiency, what timeframe should we be looking at?
00:55:46Very short-term or long-term?
00:55:48Well, we look at both.
00:55:49So, we're looking at short-term efficiencies are where we've identified potential areas of concern and made recommendations of how to improve those areas of concern.
00:55:58And then, of course, with the long-term one, that gets into the system that you're using, a systemic challenge.
00:56:04And we identify opportunities there as well, sometimes as matters for Congress, sometimes as areas of attention for the executive branch.
00:56:11So, we look at both of them.
00:56:13I want to ask Mr. Canarsa that same question.
00:56:15Yeah, that was a good answer.
00:56:16OIGs are in the same business.
00:56:18You know, there are short-term benefits and then, you know, project out for the longer term for those bigger efficiencies that you're talking about.
00:56:24And I want to follow up with one last question for you, Mr. Canarsa.
00:56:27Do you think that Inspector General's oversight, transparency, disclosure, that these are important things that we need to have real efficiency in government, whether we're using AI or not?
00:56:38Absolutely.
00:56:39That's what the, you know, the IG Act was built on, was the value of independent, transparent, nonpartisan oversight, just so that the facts can be presented as to how to improve these agencies and their programs.
00:56:51So, just to be clear, firing 19 inspectors general would not improve efficiency?
00:56:55I mean, that, I should say that the OIGs themselves didn't go away.
00:57:00Their leadership was removed.
00:57:02So, we're proud that the staff that's there and the leadership that has stepped up has kept that oversight going.
00:57:07Thank you, Mr. Chairman.
00:57:11I think, Dr. Thomas, I wanted to, you involved with government accountability offices in charge of audits, correct?
00:57:18I'm sorry, say that.
00:57:20You're in charge with, you're involved with audits of government, right?
00:57:24Yes, we are.
00:57:25Yes, okay.
00:57:26So, I'm a former auditor, spent over a decade auditing publicly traded and private companies.
00:57:31Do you think it's common sense to have authorization of transactions?
00:57:36That all transactions that government is pursued should be properly approved and authorized?
00:57:40Yes, it makes sense.
00:57:43That's right, right.
00:57:44That's what DOJ is trying to do.
00:57:45They came and said, okay, we need to make sure that transactions are properly authorized.
00:57:50This is accounting 101.
00:57:51Do you think it's proper to have justification transaction?
00:57:55What is the reason this transaction was disbursed?
00:57:57Is it based on some invoice appropriations and have all of this process?
00:58:01This is its common sense accounting.
00:58:03Yes, we support the appropriate documentation.
00:58:06Okay.
00:58:07So, that's right.
00:58:08You have documentation.
00:58:09That's what DOJ is trying to do.
00:58:10Do you think it's appropriate to have automated controls and matching of databases to make
00:58:16sure that the things when we disbursed transactions have some way of controls?
00:58:21And including maybe some automated controls.
00:58:23Do you think that's something that government should do more?
00:58:26That is a much more challenging question.
00:58:29The automation is entirely dependent upon the algorithm.
00:58:31Do you think we should have controls with manual matching, automated matching controls to match data
00:58:37to have some controls in transaction?
00:58:39Do we have to have controls?
00:58:40Do you support having controls?
00:58:41I thought your question was about automation.
00:58:43I apologize.
00:58:44Automated controls could be manual.
00:58:47I mean, but do you support controls, right?
00:58:50Do you think you're in charge of automation?
00:58:52Do you think it's sufficient technology to have more automated controls?
00:58:56I don't know if we have an answer to that right now.
00:58:58You don't know?
00:58:59So, we started a lot of controls, I don't know, 20 years ago when I was in audit, private
00:59:03companies were able to do this.
00:59:05We actually did artificial intelligence and continuous audit and done in private companies.
00:59:10Do you think government cannot do it?
00:59:13You're in charge of technology and innovation.
00:59:16How far we need to go?
00:59:17How many decades behind do you think government should be from the real world?
00:59:21Within the science domain, which is my area of responsibility, the science right now does
00:59:27not support artificial intelligence that is deterministic.
00:59:32And that is the challenge that exists for using artificial intelligence for deterministic
00:59:36choices like you're talking about.
00:59:38No, but just for simple transactions.
00:59:40You know, we're using just to identify fraud because that actually gives more efficiency.
00:59:45Don't you agree that actually continuous auditing and using more automation would detect and
00:59:51prevent fraud more efficiently than manually?
00:59:53We do like, yeah, continuous auditing is a practice that we also enjoy.
00:59:58Right.
00:59:59And I think that's what DOJ is trying to do.
01:00:01So, they try to improve control environment to avoid to have, you know, I'm not sure why
01:00:06there is an argument even about that if we try to improve.
01:00:09It seems like everyone wants to have efficiency and better auditing that the things go to proper
01:00:14causes.
01:00:15And we need to simplify.
01:00:16Don't you agree we need to simplify our billing?
01:00:19So, your first statement was about the actual technology.
01:00:24Right.
01:00:25And that is what I'm here to talk about.
01:00:27Okay.
01:00:28We do not agree, across everybody, that the existing AI technology can make error-free
01:00:36determination in an entirely automated way.
01:00:38Well, there is no one can do error-free, but that could be a tools and mechanism to help
01:00:42us, you know, to be able to detect fraud and have more controls.
01:00:47You know, there is no proof, but it's less likely to have some in-set-up system.
01:00:51So, I think that's extremely important because we spend a lot of money in healthcare, but we
01:00:56need to make sure that it's actually bring value.
01:00:58But there is a different conversation.
01:01:00It's a little bit broader, you know, not just fraud that we need to make sure that it's
01:01:04not happening, but abuse of the laws because Congress write very broad laws.
01:01:09And unfortunately, this issue became very partisan, where it used to be over billing by
01:01:14large hospitals of Medicare was a bipartisan issue, or abuses and Medicare advantage was a
01:01:19bipartisan issue, or in Medicaid, you know, Vice President Biden at the time was calling
01:01:24Ponzi scheme some of the, you know, gimmicks, you know, that the hospitals play with provided
01:01:29taxes, and suddenly it became very partisan, and now we cannot have a discussion about the
01:01:35fraud and abuse because everyone wants here to make sure that people have better benefits
01:01:40and better healthcare.
01:01:41And because we're spending so much money on healthcare, but we are not getting outcomes
01:01:46and for value.
01:01:47And Dr. Miller, you as a doctor here, what do you think would be a key issue that we should
01:01:51address to improve healthcare value?
01:01:53I would say automation to eliminate improper payments, so that way you're either saving
01:02:00money or you're redirecting that money towards other items and services for beneficiaries.
01:02:04Okay.
01:02:05Well, it seems like we should be much more bipartisan support and agreement on this issue and not
01:02:10a partisan issue because we want Americans to be healthy, because a healthy nations, you
01:02:15know, that's the only nations can be strong.
01:02:17But thank you, and I yield back.
01:02:18Thank you, Mr. Chair.
01:02:19Thank you all for being here and delighted to be at my first JAC meeting.
01:02:39So, Dr. Thomas, I want to nerd out with you a little bit.
01:02:43Before coming to Congress six and a half years ago, I was in the energy business and I built
01:02:48an AI tool for our company because the single biggest source of profit variance we always
01:02:54had was revenue variance.
01:02:55And the reason why our revenue variance was so high was because our customers, even though
01:02:58they assured us that they knew what their energy loads would be next year, they actually
01:03:01were really bad at predicting them.
01:03:03And so I built a model to go and pull in economic data and weather data and all sorts of other
01:03:07stuff.
01:03:08And good news, I'm very proud of this, is we cut our revenue variance by 90%.
01:03:13The bad news is I got all of this precision and no fundamental knowledge of what was going
01:03:17on.
01:03:18I had very good curve fitting data, but lost knowledge.
01:03:21I remember saying this to my head of engineering at the time and he said, in his wise way, that's
01:03:26the story of every advance in technology, we get more precision at the expense of knowledge.
01:03:31My GPS tells me where I am, but I don't remember how I got here.
01:03:36I asked that in the context of how you're thinking about how to design, what do you optimize
01:03:43to?
01:03:44I mean, it's one thing to say we all want efficiency, of course, that's mom and apple pie.
01:03:49But in a business context, I knew what I was optimizing to.
01:03:53In a political context, and you can take any of these that you want, but it strikes me
01:03:56that there's four things that I don't know how I do the math.
01:03:58How do you deal with political trade-offs?
01:04:00You know, if we're going to design a tariff policy, let's say, that's going to balance
01:04:04the interests of exporters and importers, that's a political question.
01:04:06That's not a math question.
01:04:08How do you deal with the fact that the federal government has unlimited downside exposure?
01:04:13You know, in my company, I was protected, my floor was protected by the bankruptcy code,
01:04:17our insurance protocols.
01:04:18The government has to be there in the worst possible case scenario, which means we need to keep resources
01:04:23around that are not used most of the time.
01:04:25How do you think about that?
01:04:28How do you deal with economic impacts outside of the government's balance sheet?
01:04:32We could save a ton of money if we defunded the police, right?
01:04:35I don't think anybody would suggest that that would be societally beneficial, whether it's
01:04:38a cop on the beat or someone at the SEC, but we know that we can save money for the government
01:04:44if we do that.
01:04:45And then, of course, how do you deal with non-economic issues?
01:04:48You know, Fermilab is right outside my district.
01:04:50I don't know what the return is on basic investment, but I'm sure glad we do it.
01:04:55On the other side of my brain, Muddy Waters used to live in my district.
01:04:59If FDR hadn't funded the WPA, hadn't sent Alan Lomax down south, we wouldn't have found
01:05:04Muddy Waters and we probably wouldn't have had the Rolling Stones.
01:05:07We could have saved money.
01:05:09So like when you design these algorithms, how do you, how do you, do you have classes
01:05:13of problems that you simply don't ask?
01:05:16Or how do you think about those when you're trying to optimize for efficiency?
01:05:21Yeah, that is a very tricky question.
01:05:23So the, and we are keenly aware and you'll notice in my written testimony, we talk about
01:05:29if you go too far in one direction, you certainly over impact small business and individuals.
01:05:36Because once you start slowing down the payment systems, people who are dependent upon timeliness
01:05:41of those payment systems who are not using it, doing improper payments, payments or fraudulent.
01:05:45They're just regular folks doing, you know, running a business or running, living their lives.
01:05:50It causes significant, you know, friction.
01:05:52And so we're trying to optimize when we do analytics to try to identify the flags we talked about.
01:05:58And that's why I'm such an advocate of high quality data.
01:06:01And you've been through this experience, you know, high quality data matters because that introduce,
01:06:05if you don't have high quality data, it introduces additional error on top of the algorithmic area that just,
01:06:10error that just exists within the process that you're using to do your modeling.
01:06:14And so trying to balance those things.
01:06:17And then we actually do sit down and think through the ethics or ethical approach of it.
01:06:20You know, what is the impact of a, you know, a false negative?
01:06:24What is the impact of a false positive?
01:06:26Of course, at GAO, we are interested in reducing fraud to zero in theory.
01:06:30So, and I know we're going to be tight on time here, but all of that makes perfect sense to me.
01:06:35It raises the next question.
01:06:37What DOGE is doing?
01:06:39Are they only using algorithms that have gone through your sort of ethical test to run traps?
01:06:45Or are they running algorithms outside of your process?
01:06:49So we have only recently been asked to look at what DOGE is doing.
01:06:54And so that is.
01:06:55So you don't know.
01:06:56We just started.
01:06:57And so we are not ready to comment on that.
01:06:59But we're.
01:07:00So to the, so to the extent DOGE is using AI tools, no one on this panel knows.
01:07:04We don't.
01:07:05What they're doing.
01:07:06That's a flag.
01:07:07To anyone on the panel, is it your view that if we identify a putative relationship, some
01:07:18opportunity for efficiency, because when I built these, when I built these algorithms
01:07:22in my company, I didn't get to just change our budget.
01:07:24I had to go to our board for approval.
01:07:26Is it your view that the executive branch can implement these changes right away without
01:07:30coming to Congress for approval?
01:07:31Does anybody maintain that position on this committee?
01:07:34I think there are plenty of tools that, yes.
01:07:40I'm just asking whether the Impoundments Act and the Rescission Control Act trump here.
01:07:45Because I think to Dr. Thomas's point, we can learn a lot from AI.
01:07:49We still need to run ethics tests on the back end.
01:07:52And it is an important question whether we think the executive branch can act without congressional
01:07:57approval to implement what they learn without even telling us what they're doing.
01:07:59Because it sounds like we don't even know what the hell DOGE is doing.
01:08:02Okay.
01:08:43Because in the old days, we were doing the bolt-ons, you know, a legacy system, bolt-on, bolt-on.
01:08:49Now some of the AI packages, I understand, can take the feed, almost write it to modern code, analyze it, break it down.
01:08:57Do you have technology you can hand to him so he can stop writing me papers and reports, and we can start moving to implementation?
01:09:05And then, Dr. Miller, if you talk to one of the two staffs, because you know we have multiple bills about moving AI in clean claims and those automations,
01:09:15we're trying to convince our brothers and sisters, particularly those with relationships to CMS,
01:09:20that if we got rid of the, you know, chase and deny sort of system and did AI match to, you know, here's the benefits, here's the doctor's notes, you know, AI scrubs, pay it.
01:09:37Help us model what the savings and efficiencies are, and I'll see you in a few minutes.
01:09:41How do I take your technology, your understanding of the technology, march it over to GAO, our actual agency,
01:10:01have them stop sending me papers about ideas for implementation,
01:10:07and actually implement, because it's not 20 years ago where it took me a two-year project
01:10:14and I had to hire IBM to do a bolt-on, on my old bolt-on, on my AS400.
01:10:20Seriously, I started on an AS400 at one time.
01:10:24Why do you laugh at that?
01:10:27But I'm really serious about this.
01:10:30The ability to take data sets today and just get going on it.
01:10:34I thought you already had certain, I don't know if they technically qualify as AI programs,
01:10:40but translational programs that will take my old server and now it's Java, now it's whatever.
01:10:46How do you help?
01:10:49I don't know that GAO particularly needs my help on this.
01:10:52I think Dr. Miller, sorry, Dr. Thompson has it well under, Thomas has it well under hand.
01:10:58But the agencies face a lot of challenges for a bunch of different reasons.
01:11:03I mean, they're all their own special creature.
01:11:08And so, you know, whether it's funding, a lot of it's expertise,
01:11:14but a lot of it is just the status quo of how things have happened in the past.
01:11:18And so they don't set up a big plan, right?
01:11:20And so I think demanding a sort of action plan is probably the first one.
01:11:26Is there a way to standardize that action plan where, okay, look, it's been pointed out,
01:11:38you know, the cost of high-end talent, our compensation schedules is one of the reasons
01:11:45we can't keep them.
01:11:47We've already learned, whether it be the Wall Street Journal series in regards to MA or some
01:11:54of the other things, sometimes the one that bathes only on occasion, but it's a dropout
01:11:59of MIT because we was already bored, has the ability to do things in hours that would take
01:12:05some of those years.
01:12:07We've all had one of those in our lives.
01:12:10I'm just trying to figure out how to move away from having another hearing, another conversation
01:12:15to actual implementation.
01:12:17A decade ago, I was positive blockchain was going to be my way to build universal databases,
01:12:29access hierarchies, permission.
01:12:32I have an expertise in blockchain node coding and design.
01:12:37And I'm here a decade later, and it didn't go anywhere.
01:12:42So help me.
01:12:43Yeah, it's a, I mean, I think it's less of a, I mean, there's a, there is obviously
01:12:49the technical problem of how do you provide these tools, but I think there is very much
01:12:53a social and political challenge here.
01:12:56And that's, that's an area that I'm not a deep expert on, although I've seen, you know,
01:13:02changes at the Federal Trade Commission primarily.
01:13:05So.
01:13:06Okay.
01:13:07Dr. Thomas.
01:13:09Yeah.
01:13:10And I'll, in this instance, if you don't mind, I probably,
01:13:13I'll spot, talk from my professional experience as opposed to GAO, because we, I don't think
01:13:19we have a body of work quite on, on this topic yet.
01:13:22You know, there's two areas I think you're getting at.
01:13:24One is the, just the workforce, right?
01:13:26And there are very bright people who are interested in working for the government because it makes
01:13:30a difference, right?
01:13:31They're mission oriented.
01:13:32You know, they don't, maybe you're not, they're at some point in their career or whether
01:13:36they're not after the big money or they've already done that.
01:13:39So that's happens because I'm one of them, you know, I have one of those people who've
01:13:42done that.
01:13:43But the second part is that getting the new technology into individual agencies one by
01:13:50one is a challenge.
01:13:51Okay.
01:13:52So can I give you, you're more than welcome to say I'm way over my skis.
01:13:57Okay.
01:13:57So I have dozens and dozens of different government entities.
01:14:01They all each have their own databases, their own mechanisms.
01:14:05If I had them basically hosted with their own sets, their own permissions, but in a common
01:14:12area, could I actually have a machine learning package that first does this one, then rolls
01:14:20to this one, then rolls to this one, and you actually do where it actually gives you your
01:14:26optionality on your data sets and externalities, you know, where it's seeing abnormal abnormalities
01:14:33and those things.
01:14:34Yeah.
01:14:34And in that case, it's a central AI shop.
01:14:40But because we're actually heading towards a world where we're going to do some more data
01:14:44sharing to find out, so like we've learned with some of the social security databases,
01:14:49the death file, because it didn't have access in so many things over here, people would steal
01:14:54grandma's identity, but grandma hasn't been with us for a while.
01:14:59And they were taking out loans.
01:15:02And that's elementary school level coding of just sharing data.
01:15:07It is the concept of, you've talked about professional skill sets and staff, but could it be in a single
01:15:13island instead of going around to the individual agencies?
01:15:16Is that a?
01:15:18It can be done.
01:15:19Now, remember, the rules for privacy of the data is defined by Congress, so that's something
01:15:24you can change.
01:15:25But look, the Defense Department keeps much of its data in an encrypted cloud.
01:15:30So yeah, the international security data is a little bit different.
01:15:32My iPhone has an encryption chip in it.
01:15:34The fact of the matter is this is more secure than any HIPAA database out there.
01:15:39Yeah.
01:15:39But as a scientist, what you're describing is a great environment where the data is there.
01:15:43The analytics are there.
01:15:44The technology exists to create role-based permissions so that you can control who accesses
01:15:52what within a centralized system.
01:15:53The technology is there.
01:15:55I think there's barriers within the government.
01:15:59A lot of it come down to the rules that have been put into place.
01:16:01Okay.
01:16:03Am I insane?
01:16:04Excuse me.
01:16:05Am I being too techno-utopian?
01:16:09You know, I think the aspiration is worth pursuing.
01:16:14I think some good points have been raised that it's about empowering the agencies, you know,
01:16:20with maybe is it dedicated resources to push forward.
01:16:24But in your vision, it's the talent is at the agency.
01:16:27And one of the conversations I've had with certain folks is saying,
01:16:30create the talent, but they capture the agency's data more from a central hub.
01:16:40I think my point is it's priorities.
01:16:44And, you know, we were talking about it briefly here, that a lot of these agencies, you know,
01:16:48they're in the business of their customer delivery.
01:16:50You know, they're trying to keep the business going.
01:16:52And it's whether it's implementing the recommendations I talked about from the OIGs
01:16:57or it's pursuing these advanced tools, it's prioritizing for them what's important,
01:17:05what's going to deliver the greatest return.
01:17:06I'm going to come back to you on that.
01:17:07Okay.
01:17:08Forgive me, but Dr. Miller is one of my favorite people,
01:17:12which, trust me, in this place doesn't get you much.
01:17:16If I came to you right now and said,
01:17:18give me the three things technology can do to change the,
01:17:22the cost and access to health care for the better for society.
01:17:30Give me the three pieces of legislation someone like me should be trying to drive through this process.
01:17:37What would you do?
01:17:39I'd focus on the plan admin processes, which are the three things that I mentioned.
01:17:44So the automation of Medicaid eligibility, because it's categorical and financial.
01:17:49There's not really discretion involved in, a lot of discretion involved in Medicaid eligibility.
01:17:54So move it towards automation.
01:17:57Require CMS to do that and have a clear timeline.
01:18:00Okay.
01:18:00So you'd have the Medicaid eligibility grab the state data and CMS do the match.
01:18:08Or the, yeah.
01:18:09Would you do a match to private data?
01:18:11To, to, to what data?
01:18:13Um, and ways and means we've talked about certain things where we have certain earned income tax credit benefits.
01:18:20So those things, instead of putting them on delays for a couple months to see if you get multiple, you know,
01:18:25requests under the same social security number, if you did a match to certain commercial databases saying,
01:18:29Oh yeah, consumption matches reported income, send them their check.
01:18:34Um, would you do a data quality bounce of a commercial database just to see if the eligibility qualifications to speed it up?
01:18:43I, I don't think it's unreasonable.
01:18:45I can't say if that would be the right operational answer.
01:18:48Okay.
01:18:48But the idea of making the standard for Medicaid eligibility, instead of it being a paper or a manual or a fax based process to have it be an automated technological process,
01:19:00just like you log on and get into your bank account.
01:19:03It should be simple.
01:19:04The second thing I would do is I would tell CMS, you could do it through oversight.
01:19:09You could also do it through legislation to move diagnosis coding across the Medicare program, not just MA,
01:19:15but also in fee for service to an automated software.
01:19:20Okay.
01:19:20And we actually have, cause, cause that's the one I was hoping you'd give me as number three,
01:19:25cause that one we're, we have great interest in.
01:19:28Cause that, that, cause the, the, the issues about diagnosis coding, right?
01:19:31There, there's clinically appropriate diagnosis coding intensity.
01:19:35There's sort of the gray area and then there's the fraud, the DOJ hauls you off to jail.
01:19:39And then there are, then there are, of course, changes that happen when beneficiaries, the same beneficiary and fee for service and MA can look differently,
01:19:48but that's cause the programs have different incentives for coding.
01:19:51So if you fix coding across the entire system, as opposed to in one of the programs and you fix it at the point of delivery,
01:19:59that's a huge improvement.
01:20:00You see, now I need to talk to you cause I need to make sure we've been actually having a discussion on MA.
01:20:06What would happen if I could get some of that coding profile?
01:20:08So I know my risk adjustment before a capitated bid, right?
01:20:15You're not bidding or you're not scoring people once a bid is our, you know,
01:20:20once services are already out there, but you do it in your initial bid and therefore it changes the incentives thing.
01:20:26I'm going to try to keep you healthier than spend my time trying to score you a sicker.
01:20:30Yeah. And, and I would say that the answer is,
01:20:33is because the beneficiary would be coded the same or very similarly in fee for service versus MA.
01:20:39Oh, because it would eliminate those incentives because.
01:20:42Okay. And number three.
01:20:43So, and number three would be making the prior authorization data submission process electronic.
01:20:50It's still like one quarter or one, one third, I think, facts and a certain.
01:20:55Okay. So universal prior, prior off, would you do it?
01:20:58Well, I would say prior off in Medicare and Medicaid managed care have the data submission.
01:21:06So I'm not even talking about the process of whether it's the same specialist reviewing that of the person who's submitting.
01:21:12I'm talking about the submission of data.
01:21:15Okay. So that'd be different than a true clean claims model.
01:21:18Right. I'm, I'm talking first, first order.
01:21:21Like we haven't even gotten submission of data to be purely electronic and integrated into point of care.
01:21:27So it should be like, if you come and see me in clinic or I see someone in clinic or in the hospital and I'm typing my note or, you know, it's my notes being dictated.
01:21:36Then at the end of the visit, I click a button and say, yes, you know, I want to send this to Cigna or whomever we can't do that right now.
01:21:43And so that's, that's a problem even before, even before the other issues in prior off.
01:21:52Only because I have people screaming at me.
01:21:55I'm supposed to be doing another.
01:21:57So when you have a quick conversation, Dr. Miller, with this young man behind me, he is freaky smart and we've been working on your number two as trying to understand,
01:22:07could we get actually the risk adjustment coding in front instead of doing it over time?
01:22:16So we actually, from day one, know what helps our brothers and sisters who are receiving, particularly, an MA benefit.
01:22:24Happy to.
01:22:24To keep them healthier.
01:22:25Gentlemen, I have a dozen questions I want to send you.
01:22:30Forgive if you get a couple notes from me.
01:22:33I really do want to think about, and I wish I could say it's my idea, but I have one of those people in my life who is truly a genius and bathing is optional.
01:22:44And that's the one who've always thought saying, look, you have this data and hundreds of different concerns through government.
01:22:53You need it to, A, talk to each other, B, if it talks to each other, then you could actually do much of its mining and set design from a single location.
01:23:06So it's not genuinely my idea.
01:23:08But I need to find out, was that just a caffeinated rambling or is it a brilliant idea?
01:23:15Dr. Thomas, before we throw the gavel.
01:23:18Yeah, just at the bug in your year, a potential model to start with is something similar to PACE, which is a centralized program for analytics to fight against fraud.
01:23:32You know, and it's bringing the data together, it's bringing the analysts together, it's a starting point, something to consider.
01:23:37Okay, you started this.
01:23:39Okay, but fraud analytics are not only crowdsourced in many ways, but they're living, they're constant.
01:23:51It's not like a single match.
01:23:54My understanding is the best ones, actually, it's constant, because there's constant, every day, there's new inputs, there's new scams, there's new, you know, new bots that, so it's not just a match, good, gone, done.
01:24:12It's just like how many of you drive with Waze, it's a living data set, and I think that's actually, it was always my blockchain model, it has to be living, and I think that becomes one of the design problems so often in government, is I hire some three people, they're very smart, they build it, but six months later, everything that's in there is already out of date.
01:24:39It's, as Dr. Miller, I believe, I may have actually gotten this from you, medical knowledge doubles every 72 days.
01:24:46But, yeah, you're right, but you create an ecosystem, similar to what exists in the cyber world, where new models and new signatures of those threats are made available to the community, as opposed to when someone discovers it, say someone discovers it in agency A,
01:25:05it can then be communicated to the rest of the community, through something like a centralized analytics center.
01:25:12Okay.
01:25:14Because you're right, people, the fraudsters are innovative too, and that's what we need to build a system for.
01:25:20Yeah, or the brain trust, oh, excuse me, Congress comes up with another idea, throws out something, and we give you absolutely unrealistic implementation times,
01:25:30and in that we create the fragilities.
01:25:35I see a whiteboard session coming.
01:25:38Gentlemen, I apologize to you.
01:25:40You deserve so much more of our attention, but what you do is important, and thank you for helping us, and with that, we're adjourned.

Recommended

3:44:00
Up next
The House Ways And Means Committee Holds Hearing To Discuss DOGE Authority And Oversight
Forbes Breaking News
21:40
Senate Democrats Hold A Press Briefing On DOGE Cuts To The Social Security Administration
Forbes Breaking News
4:53
Ayanna Pressley To Comptroller General Gene Dodaro: 'Do You Know How The Term DOGE Came About?'
Forbes Breaking News
1:36:11
Ron Wyden Leads Senate Finance Committee Hearing About Community Economic Development
Forbes Breaking News
1:48
Suhas Subramanyam Warns That DOGE Cuts May Affect Military, Other Agencies That Do 'Critical Work'
Forbes Breaking News
4:37
Shri Thanedar To GAO Official: What Effect Will DOGE's 'Sledgehammer' Approach To Downsizing Have?
Forbes Breaking News
2:46:11
Mark Green Leads House Homeland Security Committee Hearing On Artificial Intelligence
Forbes Breaking News
3:32
Judy Chu: Democrats ‘Will Not Be Submissive And Obedient’ To What DOGE Is Doing
Forbes Breaking News
1:01:28
Senate Appropriations Committee Holds A Markup On Pending Legislation
Forbes Breaking News
2:13:26
BREAKING NEWS: Jim Jordan Leads Weaponization Committee Hearing To Probe ATF Overreach
Forbes Breaking News
1:16:34
Ben Cardin Leads Senate Foreign Relations Committee Confirmation Hearing For Pending Nominees
Forbes Breaking News
1:44:25
Pete Sessions Leads House Oversight Committee Hearing About Improper Payments And Fraud At Agencies
Forbes Breaking News
2:25:31
FULL DOGE HEARING: NPR, PBS Execs Face Tough Grilling At Hearing Led By Marjorie Taylor Greene
Forbes Breaking News
5:12
Sheldon Whitehouse Presses Witnesses About The Economic Consequences Of Economic Inequality
Forbes Breaking News
4:31
Pressley Says Infrastructure Projects Must Be Done 'In A Way That Addresses Longstanding Inequities'
Forbes Breaking News
29:21
BREAKING NEWS: President Biden Holds A Briefing With Ukraine's Zelensky After G7 Agreement
Forbes Breaking News
16:26
JUST IN: House Democratic Leadership Hold A Briefing Ahead Of Biden Executive Order On The Border
Forbes Breaking News
2:50
Darin LaHood Commends Trump And Musk For Contronting Government Inefficiency ‘Head On’
Forbes Breaking News
2:53:36
House Energy & Commerce Committee Holds Hearing On CDC’s Efforts To Improve Public Trust & Health
Forbes Breaking News
2:09:30
Senate Commerce Committee Holds Hearing About Major Broadband Affordability Programs
Forbes Breaking News
57:28
Senate Armed Services Committee Holds A Hearing On Utilizing AI To Advance DoD’s Cyber Capabilities
Forbes Breaking News
2:02:16
House Committee On Energy And Commerce Holds A Hearing On 'Accusations Of Ideological Bias At NPR'
Forbes Breaking News
4:30
Mike Thompson: ‘The American People Are Fed Up With This Administration And DOGE’
Forbes Breaking News
1:41:20
BREAKING NEWS: Boeing Airline Whistleblower Testifies Before Senate Homeland Security Committee
Forbes Breaking News
51:53
JUST IN: White House Holds Press Briefing After Americans Released In Prisoner Swap With Russia
Forbes Breaking News