Congressional Testimony
Congressional Testimony
Here's a look at documents involving congressional testimony and member statements
Featured Stories
House Space & Technology Committee Chairman Babin Issues Opening Statement at Hearing on Fraud Prevention
WASHINGTON, April 23 -- Rep. Brian Babin, R-Texas, chairman of the House Science, Space and Technology Committee, released the following opening statement from an April 15, 2026, Subcommittee on Investigations and Oversight hearing entitled "Fraud Prevention: Understanding Fraud in Federally Funded Programs Run by the States":* * *
Thank you, Mr. Chairman, for holding this hearing on this important topic.
Science is one of America's greatest engines of innovation, economic growth, and national security. The driving force of that engine is our ecosystem of scientific publishing--made up of journals, ... Show Full Article WASHINGTON, April 23 -- Rep. Brian Babin, R-Texas, chairman of the House Science, Space and Technology Committee, released the following opening statement from an April 15, 2026, Subcommittee on Investigations and Oversight hearing entitled "Fraud Prevention: Understanding Fraud in Federally Funded Programs Run by the States": * * * Thank you, Mr. Chairman, for holding this hearing on this important topic. Science is one of America's greatest engines of innovation, economic growth, and national security. The driving force of that engine is our ecosystem of scientific publishing--made up of journals,academic institutions, federal agencies, and the researchers themselves. This ecosystem is undergoing rapid transformation due to evolving funding models, artificial intelligence, and increasingly competitive academic incentives. While some of these changes benefit the production and dissemination of science, others threaten productivity, innovation, and public trust.
Earlier this year, this Committee sent letters to federal science agencies expressing concerns about the growth of paper mills and predatory journals. These entities exploit the pressure to "publish or perish," which can flood the system with low-quality or even fraudulent research. When bad science gets published, it wastes taxpayer dollars, misleads policymakers, and can even put public health at risk.
But our agencies are not the only ones who must protect the scientific publishing ecosystem. Academic culture and institutions, and the publishers of scientific journals, also play major roles.
One issue is that the "publish or perish" culture prioritizes quantity over quality, creating opportunities for predatory journals and paper mills to profit while eroding scientific credibility.
Peer review, the cornerstone of scientific quality, is also under stress. Standards are inconsistent, transparency is limited, and conflicts of interest often go undisclosed. The result is research that may pass peer review but fails reproducibility tests or misguides subsequent studies.
Reproducibility and data transparency must also remain critical. Research findings are meaningful only if independent teams can verify them. Yet datasets are often inaccessible, replication studies are undervalued, and open-access policies are inconsistently enforced. A system that rewards transparency, reproducibility, and responsible data sharing strengthens public trust and ensures that science serves the American people.
Additionally, emerging technologies like artificial intelligence and machine learning can be powerful tools for discovery, but they also introduce profound risks. AI can accelerate research, analyze data, and detect errors, but it can also be misused to generate fraudulent papers or manipulate research results. Without clear oversight, these technologies could amplify the very problems we are trying to solve.
That is why we have invited our witnesses today: to discuss how the publishing industry is addressing these issues and to identify best practices in the publication of scientific research. This hearing is about risk mitigation and building trust in both the scientific community and the publishers who serve it.
Some best practices relate to retractions and corrections. This has been an area of increasing scrutiny. We are pleased to have Retraction Watch with us to discuss how retractions are a necessary part of the scientific process, as well as the damage that can occur when they are delayed, ignored, or not publicized.
Science must serve the public interest--not just the publishing industry or academic career advancement. By taking these steps, we can preserve the integrity of American science, strengthen trust in research, and ensure that taxpayer dollars are invested wisely.
I look forward to hearing from our witnesses today, and I hope today's discussion will be the first step toward policies that reinforce rigorous, transparent, and trustworthy science.
* * *
Original text here: https://science.house.gov/2026/4/opening-statement-of-chairman-chairman-brian-babin-at-the-state-of-scientific-publishing-assessing-trends-emerging-issues-and-policy-considerations
International Independent Fraud Expert Westbrooks Testifies Before House Oversight & Government Reform Subcommittee
WASHINGTON, April 23 -- The House Oversight and Government Reform Subcommittee on Government Operations released the following written testimony by Robert A. Westbrooks, international independent fraud and anticorruption expert, and former director of the Pandemic Response Accountability Committee, from an April 15, 2026, hearing entitled "Fraud Prevention: Understanding Fraud in Federally Funded Programs Run by the States":* * *
Chairman Sessions, Ranking Member Mfume, Members of the Subcommittee, I am pleased to speak with you today about fraud in federally funded, state-run programs.
My ... Show Full Article WASHINGTON, April 23 -- The House Oversight and Government Reform Subcommittee on Government Operations released the following written testimony by Robert A. Westbrooks, international independent fraud and anticorruption expert, and former director of the Pandemic Response Accountability Committee, from an April 15, 2026, hearing entitled "Fraud Prevention: Understanding Fraud in Federally Funded Programs Run by the States": * * * Chairman Sessions, Ranking Member Mfume, Members of the Subcommittee, I am pleased to speak with you today about fraud in federally funded, state-run programs. Myname is Bob Westbrooks. I am a former federal Inspector General (IG). During the national COVID-19 pandemic emergency, I served as the Executive Director of the Pandemic Response Accountability Committee (PRAC). I authored the book titled Left Holding the Bag: A Watchdog's Account of How Washington Fumbled its COVID Test about the historic level of pandemic aid fraud, the gaps in our fraud prevention and detection efforts, and how we can be better prepared before the next emergency spending crisis. I currently serve as an international independent fraud and anticorruption expert.
The following observations and opinions are based on my experiences over a thirty-year public service career, investigating fraud and auditing program fraud controls. My testimony will focus on aspects of some of the largest federally funded, state-run programs.
Summary
Millions of Americans across the nation are dependent on federally funded, state-run programs for food, housing, healthcare, temporary cash assistance, and other essential public benefits. Unfortunately, these programs like most federally funded programs are big, soft targets for fraudsters. The prevalence of fraud and the underlying root causes vary among programs. Weak controls, the emergence of AI and other technologies, and the easy accessibility of stolen personal data allow individual bad actors and international crime rings to commit fraud quickly and at scale.
Rightfully, safeguarding public monies from fraud and waste should be a national priority, and counter fraud initiatives should be supported by a coordinated, comprehensive risk-and-data-driven approach leveraging the independent Inspectors General and other accountability partners to reduce losses and help restore public trust.
Government Benefit Fraud Generally
At its core, government benefit fraud involves lying about your identity, eligibility, or claim amount.
Our pandemic experience provided a magnified view into government benefit fraud that is instructive. The internet has reduced barriers to fraud. Online, offenders can find free fraud tutorials, purchase stolen identities for the price of a Happy Meal, and file claims from anywhere in the world. With automation tools, offenders can simultaneously file multiple claims in multiple states. The prevalence of online fraud discussions normalizes this behavior and reduces the fear of getting caught and punished.
Government benefit fraud is committed by offenders across the criminal competency continuum from first-time novices who cannot resist the lure of easy money to experienced professionals, including transnational crime rings and nation-state actors.
During the pandemic, hundreds of billions in public monies were plundered by impulsive novices, experienced opportunists, and professional criminals from around the world. Our pandemic experience should have outraged Americans and spurred government to act to harden our defenses. It was all but certain that fraudsters would return to easy targets after the pandemic crisis abated. Instead, in my view, we collectively shrugged our shoulders and COVID amnesia set in.
According to the U.S. Sentencing Commission, there were a total of 937 government benefit fraud convictions in FY 2024. This represents only a tiny fraction of actual government benefit fraud as the data only includes cases identified and prosecuted to conviction at the federal level. Nevertheless, there are some insights to be gleaned from the data. For example, around 7o percent of offenders had little or no prior criminal history and the median loss was around $138,000. The top five districts for government benefit fraud were: (1) Southern District of Florida, (2) Eastern District of North Carolina, (3) District of Puerto Rico, (4) District of South Carolina, and (5) tie between Eastern District of Virginia, Southern District of New York, and Southern District of Texas.
Alabama led the nation in reported SNAP theft in the first quarter of 2025, with over 26,000 stolen benefit claims resulting in $10 million in losses. Other states with high reported SNAP benefit theft include Georgia, New York, California, and Ohio. There have been several recent SNAP cases involving Romanian nationals installing Electronic Benefit Transfer (EBT) card skimmer devices in retail stores in North Carolina, Oregon, Washington, Pennsylvania, and Mississippi.
In FY 2024, the federal government reported an improper payment estimate in the largest federally funded, state-run program, Medicaid, of $31 billion, or roughly 5 percent (some of which is fraud). The federal government reported an improper payment estimate in the Supplemental Nutrition Assistance Program (SNAP) of $10.5 billion, or roughly 11 percent (some of which is fraud). The federal government does not even estimate improper payments for the Temporary Assistance for Needy Families (TANF) program, so we have no idea how much fraud may be present in this program.
For comparison purposes, the federal government reported an improper payment estimate of $31 billion, or roughly 8 percent (some of which is fraud) in the Medicare program which is not state-run.
As the Government Accountability Office and various Offices of Inspectors General have reported, the actual amount of program fraud may be significantly higher than the reported improper payment estimates. Whatever the true amount, headlines involving large dollar fraud cases erode public trust and can adversely impact those in need.
In summary, government benefit fraud unfortunately happens and no program or state is immune. At the same time, the American public should reasonably expect that public monies are not used to pay dead people or incarcerated individuals, or to pay duplicate claims in the same state or across states, and that funds are otherwise appropriately safeguarded.
Why are Federally Funded, State-Run Programs Susceptible to Fraud?
Federally funded, state-run programs vary greatly in purpose and design but share at least two general characteristics that increase the fraud risk: (1) decentralized administration and (2) the moral hazard created when Washington provides the funding and the states control disbursement. Decentralized administration can lead to fragmentation with operational siloes and inconsistent priorities, policies, and practices.
States controlling disbursement of federal funds may diminish accountability and incentivization.
In plain language: these programs invite fraud because that's where the money is, it's relatively easy to steal, and fraudsters do not sufficiently fear getting caught and punished. In more technical jargon: these programs generally present risk factors that permit fraud, and program officials can do more to assess and treat these risks.
Fraud risk factors for these programs include, but are not limited to: program complexity, participant or payment volume, the use of third-party processors, the availability of eligibility databases, reliance on self-certification, and monitoring and oversight mechanisms.
Medicaid is by far the largest such program (approximately 10 percent of total federal spending) and is the most complex. It is an open-ended entitlement program where the federal government provides about 65 percent of total Medicaid funds and the states provide the other 35 percent. Payments are made to providers by the states either as fee-for-service or through managed care plans. Program integrity is a joint responsibility, with the Centers for Medicare & Medicaid Services (CMS) setting the standards and rules, and the states managing and monitoring the program. Dedicated State Medicaid Fraud Control Units (MFCUs) investigate and prosecute health care providers, and these MFCUs are overseen and recertified by the HHS Office of Inspector General.
SNAP is also an entitlement program. The federal government fully funds the program, but the federal and state governments share program administrative costs. H.R. 1 made significant changes to SNAP's funding structure based on reported payment error rates.
Beneficiaries typically submit applications online, by mail, or in person to their local departments of social services. Benefits are paid monthly via an EBT card to purchase eligible food items from authorized retailers. State offices are responsible for investigating fraud by recipients. The USDA Office of Inspector General is responsible for investigating retailer fraud and trafficking. Oversight is supported through the SNAP National Accuracy Clearinghouse, which is an interstate data matching system to prevent participants from receiving benefits in multiple states. According to public data, 15 states are currently using this system. There is a regulatory deadline of October 2027 for all states to implement this system.
The TANF program is a fixed, federal block grant with state matching requirements. The program is intended to provide temporary cash assistance, child care, and job training.
Because TANF is a block grant, these funds come with fewer federal requirements and minimal federal oversight.
The unemployment insurance (UI) program is primarily financed through employer payroll taxes, with the U. S. Department of Labor (DOL) ensuring that states comply with federal law and maintaining a trust fund which, among other things, reimburses states for the federal share of any extended benefits in periods of high or rising unemployment. Fraud cases are investigated by state authorities and the DOL Office of Inspector General. Program integrity efforts are supported by the UI Integrity Center, which is a partnership between the DOL and the National Association of State Workforce Agencies. Among other things, the Center's Integrity Data Hub provides cross-matching of claims in more than one state and maintains a suspicious actors repository.
The Housing Choice Voucher program provides rent subsidies in the form of vouchers based on household income. This Department of Housing and Urban Development (HUD) program is administered through local public housing authorities. To reduce red tape on initial enrollment for applicants experiencing homelessness, applicants are sometimes permitted to self-certify identity, income, and family composition. Housing Choice Voucher fraud is investigated by state and local authorities and the HUD Office of Inspector General.
In summary, fraud risk factors vary greatly among federally funded, state-run programs.
What Can and Should be Done to Reduce Fraud
In an ideal world, agencies would design fraud out of the program before implementation to the extent practicable and then would employ reasonable controls to: (i) deter fraudsters from even attempting a fraudulent claim, (ii) prevent fraudsters from consummating the claim, (iii) promptly detect those claims that successfully pass, and (iv) correct/respond to fraudulent claims through prompt enforcement actions, target hardening, disruption, and recovery.
There is no one-size-fits-all solution. Officials must consider relevant fraud risk factors and weigh program objectives and the operating environment to design fit-for-purpose fraud controls. For example, in the earliest days of the COVID-19 pandemic, we were facing a public health and economic catastrophe. The "pay-and-chase" model is not preferred in the normal course of business, but it may be the least bad choice in a crisis to ensure swift delivery of public services to the greatest number of eligible citizens. One fraud lesson from the pandemic era was that government officials, in general, did not balance this choice with stronger detection and correction controls.
Today's environment presents similar trade-off decisions for program officials. For example, an initial application for housing assistance by an individual experiencing homelessness should be addressed with different fraud controls than a recertification application.
In terms of the reasonableness of fraud controls, it should be noted that fraudsters can experiment with AI and other imperfect technologies and make mistakes along the way.
The government does not always have that luxury in the administration of public service programs. Identity theft is all too common, and a false positive regarding a SNAP or UI application can result in a family going hungry. The Dutch government learned this lesson the hard way in a widely reported scandal where a faulty algorithm resulted in wrongfully denied benefits. Similarly, in the U.S., the Electronic Privacy Information Center (EPIC) filed a complaint with the Federal Trade Commission regarding the implementation of an AI tool by some states that, according to EPIC, erroneously flagged certain public benefit claims as fraudulent resulting in the wrongful reduction or denial of benefits. A false positive result may be of little consequence to a vendor or the government, but it surely is felt by one falsely accused of fraud. Public officials should aggressively but responsibly adopt new technology tools in the fight against fraud.
To be frank, "zero tolerance" sounds great, but in reality it is simply impossible or impracticable to design a 100 percent fraud-proof program. The cost of controls and impact on the delivery of public benefits would be too great. This is not to suggest that we should "tolerate" fraud. In practice, zero tolerance means that the American public should expect that government officials are taking all reasonable steps to prevent, detect and investigate allegations of fraud.
Finally, I would be remiss if I did not briefly comment on the proposed funding cuts for some Offices of Inspectors General. USDA OIG's budget submission, for example, calls for an approximate 14 percent decrease from FY 2026. Supporting the work of the independent Inspectors General in word and deed is essential.
Illustrative Fraud Cases
In Minnesota, over 60 individuals have been convicted in a $250 million grant fraud scheme to exploit a federally funded child nutrition program. The defendants fraudulently claimed to be serving meals to thousands of children a day.
In Mississippi, the former Executive Director of the state Department of Human Services and others were convicted of misusing millions of dollars in TANF funds.
According to a report by the state auditor, former NFL quarterback Brett Favre received $1.1 million for speaking engagements he did not attend (Mr. Favre repaid the funds and has not been criminally charged), and other funds were diverted to build a new volleyball stadium at the University of Southern Mississippi.
In Arizona, a defendant from Pakistan was indicted, along with others, in connection with an alleged $650 million Medicaid fraud scheme involving at least 41 substance abuse treatment centers. The defendant owned a billing company for treatment centers and was involved in billing for services never provided and billing for substandard services. The defendant personally received at least $25 million and used some of the funds to purchase a $2.9 million home on a golf estate in Dubai.
In Florida, the owner of a network of assisted living facilities and skilled nursing facilities was convicted of fraud in what was described at the time by the Department of Justice as the "largest health care fraud scheme ever charged." The defendant bribed doctors and regulators, provided poor care to patients, and billed Medicaid and Medicare for unnecessary services. He personally received over $36 million from the scheme.
In New York, a USDA employee and others were charged in a multimillion-dollar SNAP fraud scheme that generated over $66 million in unauthorized transactions. The case involved supplying unauthorized EBT terminals to co-conspirator retailers.
Conclusion
Federally funded, state-run programs offer a broad attack surface and are susceptible to fraud. Hoping that fraudsters will not target a program is not an effective counter-fraud strategy and appealing to the better angels of our nature to not grab easy money is not an effective fraud control. We need to narrow the opportunity for fraud and make it harder to steal. Oversight is a team sport. We need a comprehensive, coordinated risk-and-data driven approach with federal and state officials, along with Inspectors General, the GAO, and other accountability partners.
Thank you for the opportunity to share my perspectives. I look forward to your questions.
* * *
Original text here: https://oversight.house.gov/wp-content/uploads/2026/04/Westbrooks-Written-Testimony.pdf
House Space & Technology Committee Chairman Babin Issues Opening Statement at Hearing on Energy, Mineral Dominance
WASHINGTON, April 23 -- Rep. Brian Babin, R-Texas, chairman of the House Science, Space and Technology Committee, released the following opening statement from an April 16, 2026, hearing entitled "Subsurface Science and Technology: American Energy and Mineral Dominance":* * *
Thank you, Chairman Weber.
Good morning. Today's hearing on subsurface science and technology comes at an important moment. While the United States has long led in innovation, the Chinese Communist Party (CCP) is using every tool at its disposal to not only surpass us but also to set the course for future innovation and ... Show Full Article WASHINGTON, April 23 -- Rep. Brian Babin, R-Texas, chairman of the House Science, Space and Technology Committee, released the following opening statement from an April 16, 2026, hearing entitled "Subsurface Science and Technology: American Energy and Mineral Dominance": * * * Thank you, Chairman Weber. Good morning. Today's hearing on subsurface science and technology comes at an important moment. While the United States has long led in innovation, the Chinese Communist Party (CCP) is using every tool at its disposal to not only surpass us but also to set the course for future innovation andstandards in subsurface science and technology.
To maintain our lead, the U.S. must continue to unleash the power of our subsurface, from energy production to establishing a critical minerals supply chain and supporting cutting-edge science. Energy production and critical minerals are increasingly intertwined. To design and operate next-generation energy production, critical minerals will play a vital role, while mining and processing those minerals will require increased energy production. Simply put, these sectors rely on each other and must move hand in hand to allow the U.S. to lead the world in innovation.
Since his first term, President Trump has worked to address the CCP's monopoly over several rare earth elements and other critical minerals. In 2017, President Trump signed Executive Order 13817, the Federal Strategy to Ensure, Secure and Reliable Supplies of Critical Minerals, which designated 35 minerals as critical to the U.S. Furthermore, he signed five presidential determinations in 2019 that declared reliance on foreign nations for rare earth elements to be a national security threat.
In his second term, President Trump has continued to drive America to build a domestic supply chain for these minerals, working with the private sector to provide the necessary tools for these companies not only to get off the ground, but also to become successful businesses for decades to come.
Cultivating and discovering new ways to access domestic critical minerals, including rare earths, should be a bipartisan issue. If the U.S. continues to rely solely on a single source, we can look to the effects of Chinese export restrictions on Japan to see the immense ramifications that could await us. This can be prevented by establishing a domestic critical minerals industry. Rare earths are essential to the U.S. defense industry and a key reason we field the world's most advanced military. Losing access would put not only the American military but also the American public in an unsafe position.
Thankfully, the Department of Energy and the National Science Foundation have been key allies in advancing innovation in critical minerals, including rare earths. Through research supported by DOE and NSF, it was found that coal ash--something we have plenty of from decades of burning coal--has the potential to be an economically viable source of rare earth elements.
While it is essential to strengthen the critical minerals supply chain to secure our economic future, we must also ensure that our power demands continue to be met. The news out of the enhanced geothermal space has been nothing short of remarkable and a reminder of what innovation in one area can lead to in another. Adding additional baseload energy sources to our grid right now is critical, and enhanced geothermal seems well-positioned to harness this resource and deliver it to the grid.
Including geothermal and critical minerals in the titles of DOE offices shows how important these two priorities are to President Trump and Secretary Wright. I look forward to discussing how these organizational changes have impacted the essential work across these sectors, and what Congress can do to support continued progress.
I want to thank our witnesses for their testimony today, and I am eager to get this conversation underway.
Thank you, Chairman Weber. I yield back the balance of my time.
* * *
Original text here: https://science.house.gov/2026/4/opening-statement-of-chairman-brian-babin-at-subsurface-science
House Science, Space & Technology Subcommittee Chairman McCormick Issues Opening Statement at Hearing on Fraud Prevention
WASHINGTON, April 23 -- Rep. Rich McCormick, R-Georgia, chairman of the House Science, Space and Technology Subcommittee on Investigations and Oversight, released the following opening statement from an April 15, 2026, hearing entitled "Fraud Prevention: Understanding Fraud in Federally Funded Programs Run by the States":* * *
Good morning. Thank you all for being here today.
We are addressing a challenge that sits at the heart of American scientific leadership--a challenge that demands honest scrutiny of the systems we rely on to produce and validate knowledge.
Science advances through trust. ... Show Full Article WASHINGTON, April 23 -- Rep. Rich McCormick, R-Georgia, chairman of the House Science, Space and Technology Subcommittee on Investigations and Oversight, released the following opening statement from an April 15, 2026, hearing entitled "Fraud Prevention: Understanding Fraud in Federally Funded Programs Run by the States": * * * Good morning. Thank you all for being here today. We are addressing a challenge that sits at the heart of American scientific leadership--a challenge that demands honest scrutiny of the systems we rely on to produce and validate knowledge. Science advances through trust.Everything we do rests on the assumption that the underlying work is real, rigorous, and honestly reported.
I know the importance of this firsthand. From my time at med school to practicing in the emergency room, my career has been based on published research and its credibility.
Our actions are based on trust -- be it when a policymaker acts on published evidence, or when a federal agency directs billions in funding to specific research.
That foundation is under threat, and it is the job of this Committee to understand why.
The scientific publishing ecosystem has changed dramatically.
What was once a straightforward process of peer review and dissemination has become a complex, commercialized marketplace with misaligned incentives and bad actors willing to exploit them.
Now consider the scale of the problem. Estimates suggest that over 400,000 published studies worldwide may originate from so-called paper mills--operations that produce fabricated or manipulated manuscripts for a fee.
In 2023 alone, Wiley, a prominent American publisher, retracted over 8,000 fraudulent papers from a single subsidiary.
That is not an anomaly-- it's a symptom. The incentive structures driving this are well understood, but under-addressed.
Academic careers are built on publication counts. Institutions compete on research output.
The result is a "publish or perish" culture that rewards quantity over quality--and creates a ready market for shortcuts.
When speed and quantity displace rigor and reproducibility, it is not just individual researchers who are harmed. It is the integrity of the scientific record itself.
Federal taxpayer dollars support the research that feeds this system.
If fraudulent publications are being cited in grant applications that shape the direction of federal policy, American taxpayers will end up subsidizing the corruption of the very science they are paying to advance.
Meanwhile, the Chinese Communist Party has built an academic incentive structures that has generated an industrial-scale paper mill problem.
Chinese universities have reportedly paid authors substantial cash rewards for publications in prestigious journals.
A survey of medical residents at hospitals in southwest China found that nearly half admitted to buying papers, selling papers, or hiring ghostwriters to meet publication requirements.
Critically, this fabricated research does not stay in China--it enters Western journals, shapes citations, and contaminates the global scientific literature on which researchers depend on.
The rise of artificial intelligence intensifies all of this.
AI lowers the cost and technical barrier to generating plausible-looking scientific content at scale.
The same tools that can accelerate legitimate research can be weaponized to fabricate data, manufacture citations, and flood the publication pipeline with content that appears credible but is not.
As these capabilities grow, the gap between what looks like science and what is science will only widen--unless institutions, publishers, and federal agencies act deliberately to close it.
Open access mandates, which this Committee has engaged with extensively, add another layer of complexity.
The movement to make federally funded research freely available is a worthy goal, but the article processing charge model used to make this financially feasible creates its own distortions.
This model incentivizes publishers to prioritize quantity over quality, disadvantaging smaller institutions, and enabling predatory journals that collect fees while providing no meaningful peer review.
The agencies under this Committee's jurisdiction have begun responding.
Following letters sent by this Committee earlier this year, agencies are developing award conditions that can lead to suspension or termination of funding when integrity requirements are not met.
That is a positive development, and we want to understand how those efforts are progressing.
But agency action has limits, and the structural pressures faced by this industry require a more comprehensive response.
This Committee's responsibility is to ensure that federal investments in science are protected, that the research enterprise maintains the credibility it requires to function, and that American institutions are not disadvantaged by competitors who treat scientific integrity as optional.
That is the purpose of today's hearing. Our witnesses can help us understand both the depth of the problem and what meaningful solutions look like.
I look forward to a serious and substantive discussion.
* * *
Original text here: https://science.house.gov/2026/4/opening-statement-of-chairman-rich-mccormick-at-state-of-scientific-publishing
GAO Director Bagdoyan Testifies Before House Oversight & Government Reform Subcommittee
WASHINGTON, April 23 -- The House Oversight and Government Reform Subcommittee on Government Operations released the following written testimony by Seto Bagdoyan, director of forensic audits and investigative services at the Government Accountability Office, from an April 15, 2026, hearing entitled "Fraud Prevention: Understanding Fraud in Federally Funded Programs Run by the States":* * *
Chairman Sessions, Ranking Member Mfume, and Members of the Subcommittee:
I appreciate the opportunity to discuss fraud against federally funded, state-administered programs and the reasons for continued ... Show Full Article WASHINGTON, April 23 -- The House Oversight and Government Reform Subcommittee on Government Operations released the following written testimony by Seto Bagdoyan, director of forensic audits and investigative services at the Government Accountability Office, from an April 15, 2026, hearing entitled "Fraud Prevention: Understanding Fraud in Federally Funded Programs Run by the States": * * * Chairman Sessions, Ranking Member Mfume, and Members of the Subcommittee: I appreciate the opportunity to discuss fraud against federally funded, state-administered programs and the reasons for continuedchallenges in combating such fraud.
The federal government is one of the world's largest and most complex entities, spending trillions of dollars across a broad array of programs and operations. The size, scope, and complexity of the federal government create inherent risks that need to be recognized and managed properly.
Fraud is one such risk that must be better managed to ensure that program delivery and taxpayer dollars are safeguarded.
State agencies administer several federal benefits programs, making payment, eligibility, and other decisions. For example, states administer Medicaid and Unemployment Insurance (UI). In addition, the federal government provides funding to state and local governments through grants, funding a wide range of public policy initiatives. In fiscal year 2025, federal grant funds totaled an estimated $1.2 trillion to state and local governments.1 Decentralized program delivery of federal programs, such as through distributed payment and eligibility decisions, can heighten the risk of fraud.
Most federal spending, including that administered through states, is not lost to fraud. However, every dollar or resource that is diverted to fraudsters hinders the federal government's ability to achieve its goals.
Direct financial losses from fraud place an increased burden on the government's financial outlook.2 Additionally, nonfinancial impacts and losses erode public trust in government and hinder agencies' efforts to execute their missions and program objectives effectively and efficiently.
While it is impossible to eliminate fraud completely, managing the risk strategically by implementing preventive, detective, and response controls is imperative. And prevention is key--attempting to investigate and prosecute our nation's way out of the problem addresses only a small fraction of fraudulent activity, requires significant time and resources, and returns pennies on the dollar. Further, fraudsters' tactics are ever evolving and so should the U.S. government's approach. To be clear, the task of managing fraud risks is never-ending. The goal is to continuously improve antifraud efforts--through analytics, evaluation, and culture--to more efficiently and effectively prevent fraud upfront, before the loss or compromise occurs.
* * *
1 Congressional Research Service, Federal Grants to State and Local Governments: Trends and Issues, R40638, (Washington, D.C.: June 26, 2025).
2 We have previously reported that the federal government faces an unsustainable, longterm fiscal future. GAO, The Nation's Fiscal Health: Strategy Needed as Debt Levels Accelerate, GAO-25-107714 (Washington, D.C.: Feb. 5, 2025). We have also reported improved efforts to combat fraud, with an emphasis on prevention, can reduce the loss of federal dollars and help improve the federal government's fiscal outlook.
* * *
This statement focuses on fraud in federally funded, state-administered programs by (1) outlining the scope of the problem, the federal fraud risk landscape, and fraud risks in federal programs administered by states; (2) examining challenges facing federal and state agencies in combating fraud; and (3) examining fraud threats the government faces and why it is difficult to combat them.
This statement is based on a body of work of selected reports we published from July 2010 to March 2026 addressing fraud risk management.3 More detailed information on the scope and methodology of our prior work can be found within the individual reports on which this statement is based. These reports are listed in the Related GAO Products page at the end of this statement. We conducted the work on which this statement is based in accordance with generally accepted government auditing standards. Those standards require that we plan and perform the audit to obtain sufficient, appropriate evidence to provide a reasonable basis for our findings and conclusions based on our audit objectives. We believe that the evidence obtained provides a reasonable basis for our findings and conclusions based on our audit objectives. We conducted our related investigative work on which parts of this statement are based in accordance with standards prescribed by the Council of the Inspectors General on Integrity and Efficiency.
* * *
3 As of April 2026, we have made 215 recommendations to federal agencies and programs to better manage fraud risks, of which about 40 percent remain open.
* * *
Scope of Fraud in Federal Programs and Risks to Programs Administered by States
No area of the federal government is immune to fraud. All federal programs and operations are inherently at risk, regardless of whether they provide financial or nonfinancial benefits or whether delivery takes place at the federal, state, or local level.4 Decentralized program delivery and distributed control environment for programs that are federally funded and state administered pose further challenges to managing fraud risks.
Federal Fraud Risk Landscape
The federal government is a major target for fraud. Fraud schemes are wide-ranging and occur when something of value is obtained through willful misrepresentation. For example, receiving a benefit or assistance from the federal government by lying to meet eligibility requirements is a fraudulent act. Fraud can include financial and nonfinancial benefits, which can have both financial and nonfinancial impacts (see fig. 1). [View figure in the link at bottom.]
* * *
Figure 1: Examples of Financial and Nonfinancial Benefit Fraud
* * *
4 Two large federal programs administered by states--Medicaid and Unemployment Insurance--are included on GAO's High Risk List. The High Risk List highlights federal programs and operations that we have determined are in need of transformation. It also names federal programs and operations that are vulnerable to waste, fraud, abuse, and mismanagement. We update our High-Risk List every 2 years at the start of each new Congress. GAO, High-Risk Series: Heightened Attention Could Save Billions More and Improve Government Efficiency and Effectiveness, GAO-25-107743 (Washington, D.C.: Feb. 25, 2025).
* * *
Understanding the scope of the problem and the impacts of both financial and nonfinancial benefits fraud are critical to combating it in government programs and operations.
In this regard, fraud risk increased significantly during the COVID-19 pandemic, when the federal government provided about $4.5 trillion in relief funds for response and recovery efforts. The emergency environment and the need to distribute the funds quickly provided increased opportunities for fraud against federal programs, including those that are state administered.
Financial fraud loss. In 2024, we estimated that the federal government loses between $233 billion and $521 billion annually to fraud, based on data for fiscal years 2018 through 2022.5 The range represents 3 percent to 7 percent of average federal obligations during that period.6 The width of the range is a reflection of both the uncertainty associated with estimating fraud and the diversity in the risk environments that were present in fiscal years 2018 through 2022. Given the time frame of the data, the estimate includes pandemic-related spending. The estimate also captures losses that occur at the state, local, tribal, or other government level if those losses included a federal investigative, administrative, or related action.7
Other estimates of pandemic fraud loss that we and the oversight community developed further support the finding of higher fraud risk in pandemic spending.8 One such fraud estimate we developed is for UI programs that provided federal economic relief through state workforce agencies. Based on statistical sampling and imputation techniques, we estimated that the amount of fraud in UI programs during the COVID-19 pandemic was likely between $100 billion and $135 billion. This is about 11 percent to 15 percent of the total amount of UI benefits paid during the pandemic.9
* * *
5 GAO, Fraud Risk Management: 2018-2022 Data Show Federal Government Loses an Estimated $233 Billion to $521 Billion Annually to Fraud, Based on Various Risk Environments, GAO-24-105833 (Washington, D.C.: Apr. 16, 2024).
6 These percentages should not be applied at the agency or program level.
7 The upper range of the estimate is associated with higher-risk environments like we saw in the pandemic, whereas the lower end reflects lower risk environments.
* * *
Nonfinancial impacts. Nonfinancial impacts may not pose a direct financial cost but can lead to other potentially harmful outcomes. We have reported that the federal government faces many such risks, including falsified passport and trademark applications, aircraft registrations, and supporting documentation for immigration-related benefit programs cases.10
* * *
8 We estimated pandemic unemployment insurance fraud to be between $100 billion and $135 billion. GAO, Unemployment Insurance: Estimated Amount of Fraud during Pandemic Likely Between $100 Billion and $135 Billion, GAO-23-106696 (Washington, D.C.: Sept. 12, 2023). The Small Business Administration Office of Inspector General (OIG) estimated fraud in two pandemic programs to be about $200 billion. Small Business Administration, COVID-19 Pandemic EIDL and PPP Loan Fraud Landscape, White Paper Report 23-09 (June 27, 2023). The Pandemic Response Accountability Committee (PRAC) estimated that three pandemic-relief programs disbursed approximately $79 billion in potentially fraudulent payments due to the use by applicants of over 1.4 million potentially stolen or invalid Social Security numbers. These programs were the Small Business Administration's COVID-19 Economic Injury Disaster Loan program and the Paycheck Protection Program, and the Department of Labor's pandemic-related unemployment insurance programs. PRAC, Fraud Prevention Alert: Pre-Award Vetting Using Data Analytics Could Have Prevented Over $79B in Potentially Fraudulent Pandemic Relief Payments (Washington, D.C.: June 2025).
9 GAO's estimate is for the period from April 2020 (first full month of payments from all UI programs) to May 2023 (end of the public health emergency). This estimate covers all 53 states that participated in the regular and temporary UI programs. The 53 number includes the 50 states, the District of Columbia, Puerto Rico, and the U.S. Virgin Islands. See GAO-23-106696.
10 See, for example, GAO, State Department: Undercover Tests Show Passport Issuance Process Remains Vulnerable to Fraud, GAO-10-922T (Washington, D.C.: July 29, 2010), State Department: Pervasive Passport Fraud Not Identified, but Cases of Potentially Fraudulent and High-Risk Issuances Are under Review, GAO-14-222 (Washington, D.C.: May 1, 2014), and Intellectual Property: Stronger Fraud Risk Management Could Improve the Integrity of the Trademark System, GAO-24-106533 (Washington, D.C.: Mar. 14, 2024).
* * *
Fraud Risks in Federal Programs Administered by States
There are many reasons and benefits to having federal programs administered through states.11 However, this approach exponentially increases the scale of government transactions across the states and U.S. territories. The programs can vary in size, but some, such as Medicaid, involve tens of millions of beneficiaries. When payment or eligibility decisions are made outside of federal agencies, fraud risk heightens.12
For example, the Council of the Inspectors General on Integrity and Efficiency reported in January 2021 that grant programs--such as the Temporary Assistance for Needy Families (TANF) block grant that is funded by the U.S. Department of Health and Human Services and administered by states--faced an increased risk of fraud, waste, and mismanagement because of limited visibility and control over expenditures at the award recipient and subrecipient levels.13 Additionally in May 2020 and October 2021, the Mississippi State Auditor announced that multiple individuals affiliated with that state's TANF program potentially misspent, converted to personal use, or wasted more than $77 million of TANF grant funds.14
Decentralized program delivery--where federal funds are distributed to grantees, subrecipients, contractors, and subcontractors--also creates vulnerabilities to different types of fraud. For example, in May 2021, we found the decentralized environment makes the Community Development Block Grant Disaster Recovery (CDBG-DR) vulnerable to certain types of fraudulent schemes as money flows from the Department of Housing and Urban Development to several entities, including states, before reaching their intended beneficiaries.15 These schemes include contractors providing false certification of qualifications or eligibility, fraudulently billing after taking a deposit and receiving CDBG-DR funds, and conspiring to influence the procurement process to circumvent competitive bidding controls.
* * *
11 Congressional Research Service, Federal Grants-in-Aid Administration: A Primer, R42769 (Washington, D.C.: Mar. 5, 2026).
12 GAO, COVID-19: Insights and Actions for Fraud Prevention, GAO-24-107157 (Washington, D.C.: Nov. 14, 2023).
13 Council of the Inspectors General on Integrity and Efficiency, The IG Community's Joint Efforts to Protect Federal Grants from Fraud, Waste, and Abuse (Jan. 2021). State TANF agencies are grant award recipients. TANF subrecipients can include local agencies, nonprofit organizations, and other contracted organizations.
14 Mississippi Office of the State Auditor, Press Releases (May 4, 2020; Oct. 12, 2021). See also, GAO, Temporary Assistance for Needy Families: Additional Actions Needed to Strengthen Fraud Risk Management, GAO-25-107290 (Washington, D.C.: Jan. 28, 2025).
15 GAO, Disaster Recovery: HUD Should Take Additional Action to Assess Community Development Block Grant Fraud Risks, GAO-21-177 (Washington, D.C.: May 5, 2021).
* * *
Federal and State Agencies Have Faced Challenges Combating Fraud in Federally Funded Programs
Challenges Faced in Demonstrating Commitment to Combating Fraud
Widespread recognition of the value of fraud risk management and commensurate commitment have been lacking at federal agencies, including those responsible for state administered programs. As we note in GAO's A Framework for Managing Fraud Risks in Federal Programs (Fraud Risk Framework), commitment to fraud risk management is a fundamental first step.16 Commitment--by program managers at federal and at state levels when administering federal programs--means creating an organizational culture and structure conducive to fraud risk management.
* * *
16 GAO, A Framework for Managing Fraud Risks in Federal Programs. GAO-15-593SP (Washington, D.C.: July 28, 2015). The framework provides a comprehensive set of leading practices for combating fraud in a strategic, risk-based way, with a focus on prevention.
* * *
Commitment to fraud risk management must start with leadership, setting the tone at the top that acknowledges fraud risks, commits attention and resources to manage them decisively, and communicates the value of fraud risk management (see sidebar). The objective of fraud risk management is to ensure program integrity by continuously and strategically mitigating the likelihood and impact of fraud. This objective is meant to facilitate achievement of the program's broader mission and strategic goals by helping to ensure that funds are spent effectively, services fulfill their intended purpose, and assets are safeguarded.17 Potential reputational fears that may result in sweeping incidents of fraud "under the rug," are, in fact, the reasons for managing fraud risks proactively. The impacts of unmanaged fraud risks on programs' reputations and the negative publicity can be severe.
Recommendations to designate an entity to lead fraud risk management were among the five key areas of needed actions, according to our analysis of recommendations we made to federal agencies on fraud risk management from July 2015 through December 2022.18 Our prior work has shown that when agencies formally designate an entity to design and oversee fraud risk management activities, their efforts can be more visible across the agency, particularly to executive leadership. Commitment to fraud risk management is fundamental to sustaining progress in other areas, such as conducting fraud risk assessments, designing effective controls, and continuously monitoring and evaluating antifraud efforts.
Challenges Faced in Various Practices for Managing Fraud Risks
The Fraud Risk Framework also discusses other leading practices for managing fraud risks. More specifically, as part of delivering on program mission, officials should develop robust fraud risk management programs at federal and state levels to manage fraud risks proactively, strategically, and continuously. This involves effective approaches in designing fraud risk management activities, such as controls and data analytics, as well as staff capacity to carry out antifraud efforts based on current knowledge and skillsets.
* * *
17 GAO-15-593SP.
18 The five key areas in which federal agencies need to take additional actions are (1) designating an entity to lead fraud risk management, (2) assessing fraud risks, (3) designing and implementing an antifraud strategy, (4) using data analytics to manage fraud risks, and (5) managing fraud risks in emergencies. GAO, Fraud Risk Management: Key Areas for Federal Agency and Congressional Action, GAO-23-106567 (Washington, D.C.: Apr. 13, 2023).
* * *
Five Principles of Fraud
* There is always going to be fraud. It is a fact that some individuals will look to gain where there is opportunity. Organizations need robust processes in place to prevent, detect, and respond to fraud and corruption.
* Finding fraud is a good thing. If you do not find fraud, you cannot fight it. This requires a change in perspective so the identification of fraud is viewed as a positive and proactive achievement.
* There is no one solution. Addressing fraud needs a holistic response incorporating detection, prevention, and response, underpinned by a strong understanding of risk. It also requires cooperation and collaboration between organizations.
* Fraud is ever changing. Fraud and counter fraud practices evolve very quickly, and organizations must be agile and change their approach to deal with these evolutions.
* Prevention is the most effective way to address fraud. Preventing fraud reduces financial loss and reputational damage. It also requires fewer resources than an approach focused on detection and recovery.
Source: International Public Sector Fraud Forum, Guide to Managing Fraud for Public Bodies, February 2019. |
* * *
However, programs have too long relied on the costly and ineffective "pay-and-chase" model, which refers to the practice of attempting to recover funds after payments have been made. This approach was particularly evident during the COVID-19 pandemic for programs administered at the federal level, such as the U.S. Small Business Administration (SBA)'s programs to support small businesses and retain employees, and for programs at the state level to support individuals, such as the Department of Labor's (DOL) UI programs.
Further, weak control environment, data and system limitations, and limited capacity to manage fraud risks have challenged the practices of federal and state program managers to manage fraud risks and build robust programs.
Weak control environment. We have found that federal and state agencies relied on self-attestation or self-certification for individuals and entities to verify their eligibility or identity to receive assistance from some COVID-19 relief programs, in an effort to disburse funds quickly to those in need.19 Even when program design decisions allow for self-certification, agencies are responsible for designing and implementing control activities to prevent fraud.
Self-certification alone is not sufficient as a fraud control to mitigate misrepresentation. Our prior work examining SBA's Paycheck Protection Program and COVID-19 Economic Injury Disaster Loan program fraud schemes identified (1) ineligible, nonoperating businesses that applied for, and obtained, program funds; (2) legitimate business owners misrepresenting eligibility regarding their criminal record, federal debt, or principal place of residence, among others; and (3) falsification of tax or other documents to obtain more funds.20 In these instances, recipients falsely self-certified eligibility. As we reported, other fraud controls to mitigate these misrepresentations were either not in place or were not effective for those programs.
Confirming eligibility of individuals receiving benefits, such as by confirming wage information or by verifying identity through data and other checks, are key controls to prevent fraud schemes that rely on mechanisms such as misrepresentation. However, design of such controls and choices in applying them across programs can impact their effectiveness. For example, we have reported that the U.S. Department of Agriculture's (USDA) Food and Nutrition Service (FNS)--in overseeing the implementation of the Supplemental Nutrition Assistance Program at the federal level--recommended a key control to prevent electronic benefit transfer (EBT) fraud schemes. EBT card security tools required an individual recipient to opt-in to use them, but the adoption rates were extremely low. 21 FNS officials and representatives from several stakeholder organizations said that few EBT recipients used these options because they were optional, perceived as difficult to enroll in, or inaccessible. EBT processors reported that only 5 to 10 percent of EBT recipients used these apps through which these tools were available.
* * *
19 GAO-24-107157.
20 GAO, COVID Relief: Fraud Schemes and Indicators in SBA Pandemic Programs, GAO-23-105331 (Washington, D.C.: May 18, 2023).
* * *
Data and system limitations. In 2023, we reported that our prior review of fraud risks and responses across COVID-19 pandemic programs identified challenges related to eligibility and identity, such as lack of information or data systems to confirm eligibility.22 Further, as part of our 2025 work on organized fraud groups, we found that state officials specifically cited challenges related to data limitations and program delivery.23 For example, a state official told us that each program is restricted to sharing data within the program. There are also limits to interagency information sharing. For example, one state agency administering a federally funded program reported that it is restricted by state and federal laws from sharing information with other programs in the state, such as information on individuals and the state services they use.
This restriction can limit the ability to connect fraudsters to potential fraud within and across state programs. Federal officials also told us that obtaining critical data, such as tax records to verify an applicant's identity or program eligibility, is time and resource intensive while also safeguarding the data.
In June 2022, we reported that state workforce agencies faced significant challenges in modernizing legacy IT systems for delivering UI programs.24
* * *
21 GAO, Nutrition Assistance: USDA Should Comprehensively Assess Benefit Theft Prevention Measures States Are Implementing, GAO-25-107964 (Washington, D.C.: Sept. 25, 2025).
22 GAO-24-107157.
23 GAO, Fraud Risk in Federal Programs: Continuing Threat from Organized Groups Since COVID-19, GAO-25-107508 (Washington, D.C.: Jul. 10, 2025).
24 GAO, Unemployment Insurance: Transformation Needed to Address Program Design, Infrastructure, and Integrity Risks, GAO-22-105162 (Washington, D.C.: June 7, 2022).
* * *
States consequently faced challenges in scaling aspects of their UI systems, such as applicant identity verification, to meet demand and detect improper payments. Moreover, we reported that flexibility in how states implemented DOL's Pandemic Unemployment Assistance (PUA) program allowed for variation across states in the systems states used for their PUA programs.25 For example, some states implemented the PUA program within their regular UI systems. This allowed those states to utilize some of the existing controls from those systems. While using legacy systems could allow states to leverage existing controls, if the legacy system had challenges, these could have been worsened by the increased volume of new claims during the pandemic.
In a December 2021 report, the Pandemic Response Accountability Committee (PRAC) cited a state auditor finding that the state's UI system had unresolved deficiencies that created processing issues.26 For example, increased claims during the pandemic resulted in many applicants encountering technical errors in that state. Similarly, in the same report, the PRAC cited another state auditor finding that the state's UI system was outdated and unable to detect and prevent fraudulent claims automatically.
Other states worked with contractors to set up new systems dedicated to the PUA program with new functionality not available in the systems used for their regular UI programs. Some states faced challenges when implementing new programs, however. For example, according to the December 2021 PRAC report, one state's UI system was based on a mainframe computer from the 1970s. When integrating the newly developed PUA program to operate with the outdated UI system, the state faced increasing system errors that sometimes denied eligible applicants' claims or delays in processing the claims.27
Capacity to manage fraud risks. We have previously reported that those making eligibility determinations or payment certifications may have limited experience or training. These limitations heightened the risk of fraud, waste, abuse, and other payment integrity issues across COVID-19 relief programs. For example, in June 2022, we reported that DOL officials cited new and inexperienced staff as one of the factors that provided opportunities for exploitation of UI programs and system vulnerabilities.28
* * *
25 GAO, Pandemic Unemployment Assistance: States' Controls to Address Fraud, GAO-24-107471 (Washington, D.C.: July 23, 2024).
26 Pandemic Response Accountability Committee, Key Insights: State Pandemic Unemployment Insurance Programs (Washington, D.C.: Dec. 16, 2021).
27 GAO-24-107471.
* * *
Capacity to manage fraud risks also involves the ability to balance program delivery with fraud controls. Multiple state program officials we interviewed during our 2025 work on organized fraud discussed this challenge.29 For example, state UI program managers discussed the challenge of needing to meet federal program requirements for timely benefit payments with unemployment insurance programs while also implementing fraud controls, such as identity verification. Additionally, according to a state program official, fraud impacted staff's ability to help legitimate program applicants, increasing wait times in that state.
Lastly, competing priorities, workload challenges, and resource constraints have been identified as reasons for federal agencies not conducting regular fraud risk assessments or assessing states' implementation of available tools to manage fraud risks.30
Fraud Threats the Government Faces and Why It Is Difficult to Combat Them Organized Groups
The federal government--including federally funded, state-administered programs--faces a wide variety of fraud threats. Our prior work has shown that, given the opportunity, organized criminal organizations, entities, and individuals from all walks of life have sought to defraud federal programs. Certain risk factors--such as program design, culture, and personal motivation--can also increase the risk that a program will be targeted by fraudsters.
Organized groups of individuals working together--both domestic and transnational--have defrauded public assistance programs, as was evident during the COVID-19 pandemic.31 Aided by technology, organized groups have targeted programs at a larger volume and at a greater speed than individuals. As we have reported, organized fraud harms people, programs, and society in many financial and nonfinancial ways that are worsened when carried out on a large scale.
* * *
28 GAO-22-105162; and GAO-24-107157.
29 GAO-25-107508.
30 GAO-25-107290; and GAO-25-107964.
31 GAO-25-107508.
* * *
Organized fraud groups that target government programs generally fall into three types and vary in size, structure, and participants. As discussed in our July 2025 report, the types are (1) organized criminal enterprises, (2) groups organized around a program, and (3) opportunistically organized groups (see fig. 2).32 [View figure in the link at bottom.]
* * *
Figure 2: Types of Organized Fraud Groups Targeting Government Programs
* * *
Organized criminal enterprises involve established criminal syndicates that may have other lines of illicit activity. Groups organized around a program may involve program participants and facilitators who conspire to defraud the program they are connected to. Opportunistically organized groups involve clusters of individuals who come together as opportunities for fraud arise.
These groups have used technology, program knowledge, and other means to commit fraud on a large scale against government programs.
The stages of an organized fraud scheme include methods such as gathering materials and information, defrauding programs on a large scale across multiple regions, evading detection, and laundering fraudulently obtained funds. As we have previously reported, the following examples illustrate how organized groups have defrauded federally funded, state-administered programs.
* * *
32 GAO-25-107508.
* * *
Organized criminal enterprises. Two foreign nationals and suspected leaders of an overseas-based transnational organized crime group fraudulently obtained UI benefits. In this fraud scheme, the two foreign nationals submitted multiple fraudulent applications to receive UI benefits. They both fraudulently claimed to be U.S. citizens and provided fake Social Security numbers. They used the illicit funds from unemployment benefits, along with funds they received by pawning stolen goods, and laundered these funds through wiring money to entities in another country. Their crimes included robbing $1.4 million in jewelry from residents of elderly communities. Together with their co-conspirators, the two foreign nationals received a total of approximately $32,250 in UI benefits they were not eligible to receive.
In 2023, they both pleaded guilty to conspiracy to launder monetary instruments and were sentenced to 30 and 36 months.
Groups organized around a program. In 2020, three pharmacy operators were sentenced for defrauding the USDA's Special Supplemental Nutrition Program for Women, Infants, and Children (WIC). The investigation disclosed that three individuals who operated the pharmacy purchased WIC vouchers from low-income recipients.
The pharmacy employed drivers to travel throughout a metro area and purchase high-value special infant formula WIC vouchers from recipients, which the pharmacy operators then redeemed for cash.
Opportunistically organized groups. In October 2024, the U.S. Department of Justice indicted a group for allegedly stealing over $2.4 million in Supplemental Nutrition Assistance Program (SNAP) benefits. The group used those stolen benefits to purchase large quantities of sports drinks and baby formula from websites associated with grocery stores, later reselling the goods on the black market. The group also allegedly evaded eligibility requirements by coercing beneficiaries into selling their benefits, often at a loss. The group then trafficked the illegally purchased benefits loaded onto SNAP EBT cards to third parties, who used the benefits to fraudulently purchase goods. As of April 2026, seven individuals have been sentenced, five individuals pleaded guilty and are awaiting sentencing, four are awaiting trial, and the case against one individual was dismissed.
Other Fraud Opportunists
In addition to organized fraud groups, entities in a wide variety of sectors, and individuals from all walks of life have defrauded public programs.
Based on a review of the Department of Justice's pandemic fraud cases, we identified different types of fraudsters who defrauded at least 19 different pandemic-relief programs, both federally and state-administered (see fig. 3).33 [View figure in the link at bottom.]
* * *
Figure 3: Types of Opportunists GAO Identified in Pandemic-Relief Program Fraud Cases
* * *
In 2019, a former operations director of a state child food program was sentenced in federal court to nearly 3 years in prison for defrauding the government and filing a false tax return. The USDA FNS reimbursed the program a set amount per meal served. On a monthly basis, the fraudster's organization submitted requests for reimbursement to the state, which, in turn, received funding from USDA. The fraudster submitted falsely inflated meal count forms; in some instances, they added meals to legitimate counts, and in others, they submitted forms for days when no meals were served at all.
Evolving Fraud Tactics
Efforts by federal and state officials to protect their programs from fraud are often outpaced and outmatched by fraudsters, whose schemes and use of technology are continuously evolving. During our 2025 work on organized fraud, federal and state officials told us that the increased use of stolen identities from data breaches, synthetic identities, and widespread availability of advanced technological tools, such as AI and bots, will continue to pose major challenges for officials who are continuously "one step" behind organized fraud groups.34 Officials from one agency stated that organized groups increasingly rely on obtaining large volumes of Personally Identifiable Information (PII) from data breaches, phishing attacks, and the purchase of stolen records from the dark web. An official from another agency said that organized groups may use AI and bots to facilitate fraud schemes, allowing them to file claims in rapid succession seconds or minutes apart. The widespread availability of these tools can allow individual fraudsters to operate at a level similar to organized fraud groups.
* * *
33 GAO, COVID-19 Relief: Consequences of Fraud and Lessons for Prevention, GAO-25-107746 (Washington, D.C: Apr. 9, 2025).
* * *
Chairman Sessions, Ranking Member Mfume, and Members of the Subcommittee, this completes my prepared statement. I would be pleased to respond to any questions that you may have at this time.
* * *
Original text and figures here: https://oversight.house.gov/wp-content/uploads/2026/04/Bagdoyan-Written-Testimony.pdf
DOE Critical Materials Innovation Hub Director Lograsso Testifies Before House Science, Space & Technology Subcommittee
WASHINGTON, April 23 -- The House Science, Space and Technology Subcommittee on Energy released the following testimony by Thomas A. Lograsso, director of the Critical Materials Innovation Hub at the U.S. Department of Energy Ames National Laboratory, from an April 16, 2026, hearing entitled "Subsurface Science and Technology: American Energy and Mineral Dominance":* * *
Chairman Weber and Ranking Member Ross, and Members of the Subcommittee, thank you for the opportunity to speak on the role of the Critical Materials Innovation Hub in strengthening the nation's scientific leadership, energy ... Show Full Article WASHINGTON, April 23 -- The House Science, Space and Technology Subcommittee on Energy released the following testimony by Thomas A. Lograsso, director of the Critical Materials Innovation Hub at the U.S. Department of Energy Ames National Laboratory, from an April 16, 2026, hearing entitled "Subsurface Science and Technology: American Energy and Mineral Dominance": * * * Chairman Weber and Ranking Member Ross, and Members of the Subcommittee, thank you for the opportunity to speak on the role of the Critical Materials Innovation Hub in strengthening the nation's scientific leadership, energysecurity, and supply chain resilience for critical materials.
My name is Thomas Lograsso and I serve as the Director of the U.S. Department of Energy (DOE)'s Critical Materials Innovation Hub (CMI). I am a senior leader at the U.S Department of Energy's Ames National Laboratory on the campus of Iowa State University, a single program Office of Science laboratory with the mission to deliver critical materials solutions to the nation.
I am also an adjunct professor in the Department of Materials Science and Engineering at Iowa State University.
The Critical Materials Innovation Hub is one of the DOE Energy Innovation Hubs created to accelerate scientific discovery and address critical energy issues. The Critical Materials Innovation Hub was established in 2013 in response to growing vulnerabilities in U.S. supply chains for materials essential to energy generation and use, advanced manufacturing, and national security technologies. The Hub was created following DOE assessments identifying rare earth elements and other specialty materials as increasingly subject to supply disruption, price volatility, and geopolitical risk. Today, these concerns have intensified as demand for energy related materials accelerates globally.
CMI is led by Ames National Laboratory. It operates as a coordinated national consortium of national laboratories, universities, and U.S. industry partners. This structure enables the Hub to integrate fundamental materials science, applied engineering, and manufacturing relevant research within a single, sustained framework. In 2023, DOE renewed CMI for a third five year phase of operation and formally redesignated it as the Critical Materials Innovation Hub, reflecting its maturity, its expanded mission to accelerate innovation and commercialization, and its status as a signature initiative in DOE's innovation pipeline.
Accelerating Solutions
The mission of the Critical Materials Innovation Hub is to accelerate scientific and technological solutions that enable secure, resilient, and diversified domestic supply chains for materials critical to U.S. energy technologies and manufacturing competitiveness. The Hub pursues this mission through four integrated technical pillars:
1. enhancing and diversifying supply;
2. developing substitutes;
3. unlocking secondary sources through recycling and reuse; and
4. conducting crosscutting research in thermodynamics, AI assisted predictive modeling and automated synthesis, technoeconomic and life cycle analyses, and supply chain analysis.
CMI focuses on four materials groups: rare earth elements, energy storage materials, platinum group metals, and electronic materials. Global supply chains for many of these materials remain highly concentrated. That is why long term, coordinated federal investment is required to develop alternative sources as well as to develop new materials with fewer or no critical materials, and to develop new methods to process critical materials with enhanced efficiency.
The CMI Hub provides that capability. The Hub serves as a national resource for reducing the criticality of materials across the materials life cycle.
Research, Innovation and Workforce Development
CMI's mission is pursued through a set of interrelated goals and objectives that span research, innovation, and workforce development.
A core objective is to reduce material criticality across the life cycle -- to reduce dependence on fragile or concentrated supply chains by addressing criticality at multiple points in the materials life cycle. This includes advancing technologies for diversifying primary supply, improving extraction and separation processes, and converting critical minerals into high value materials more efficiently and responsibly.
CMI is working to develop substitutes and material efficient processing. We seek to develop substitutes that meet or exceed performance requirements while using less critical or more abundant elements. This includes reduced rare earth permanent magnets, alternative functional materials, and designs that use critical materials more efficiently without sacrificing performance.
Another major goal is unlocking secondary sources of critical materials through recycling, reuse, and recovery from manufacturing scrap, mine waste, and end of life products. By improving process economics and environmental performance, CMI supports a more circular and sustainable materials economy.
We are working to advance crosscutting and predictive research. CMI conducts crosscutting research in chemistry, materials physics, materials science, environmental sustainability, and supply chain and economic analysis. These efforts provide forecasting tools and decision frameworks to anticipate which materials may become critical in the future and to guide proactive technology development.
CMI places significant emphasis on workforce development. We train students, postdoctoral researchers, and early career scientists in interdisciplinary critical materials research. To date, more than four hundred students and postdoctoral researchers have been trained through CMI, forming a national pipeline of scientists and engineers with expertise in critical materials. These products of the CMI Hub contribute to government, industry, national laboratories, and academia, strengthening long term national capacity in this strategic area.
CMI places strong emphasis on accelerating progress for the benefit of the nation. A central management objective is to accelerate innovation, shortening the transition from laboratory discovery to scalable technology and commercialization. CMI emphasizes scalable processes, pilot scale demonstrations, intellectual property generation, and close engagement with industrial partners to move technologies toward adoption in U.S. manufacturing. Through strong industry partnerships, cost shared projects, and an active intellectual property program, the Hub advances technologies toward pilot scale demonstration and early commercialization.
Since its inception, CMI has generated more than 225 invention disclosures, 72 U.S. patents, 45 licensed technologies, 17 R&D 100 Awards, and more than 750 scientific peer reviewed publications. Furthermore, CMI has attracted more than $160 million of public and private investments to move technologies into practice. These results demonstrate a significant return on federal investment.
To fully appreciate the power of the collaborative innovation ecosystem DOE has established, allow me to provide three examples of technologies that have progressed toward deployment.
Each relates to the rare earth material supply chain with emphasis on permanent magnet applications.
Success in Rare Earth Separation
Separating rare-earth elements to enable their use in many technologies and processes has been a persistent challenge. Sucessfully addressing this challenge demonstrates how sustained federal investment and coordination across the U.S. research ecosystem can translate basic scientific discovery into real-world deployment.
The effort began when scientists studying unusual microbes discovered that some bacteria naturally require rare earth elements to function. In 2018, researchers identified and characterized lanmodulin, a small protein with unprecedented affinity and selectivity for rare earth elements -- an ability far surpassing conventional chemical extractants. This discovery was notable because rare earth elements are chemically similar and traditionally require hundreds of energy- and chemical-intensive solvent extraction steps to separate. Early research demonstrated that lanmodulin could selectively bind rare earths under harsh industrial conditions, confirming its relevance beyond basic science.
The technology advanced with support from CMI, where lanmodulin evolved from a single natural protein into a suite of engineered variants tailored to specific separation challenges. This accelerated evolution was enhanced with AI methodologies to predict new variants with specific attributes. Researchers showed for the first time that protein-based systems could separate closely related rare earth elements, achieving purities exceeding 98 to 99 percent in a single processing step. These results meet performance thresholds relevant for magnet production and defense applications. They also offer the potential for significantly lowering environmental impact compared to conventional solvent extraction.
Subsequent federal support emphasized robustness, manufacturability, and pilot scale demonstration. Research focused on recovering rare earths from domestic sources such as mine waste, coal byproducts, and recycled electronics--directly supporting supply chain resilience and national security priorities. Lanmodulin-based systems were integrated into continuous, allaqueous workflows and tested well beyond laboratory scale, substantially reducing technical and commercial risk.
With scientific performance demonstrated and engineering risk mitigated, the technology transitioned to the private sector. Alta Resource Technologies licensed lanmodulin-based intellectual property and is deploying engineered protein systems for commercial and defenserelevant rare earth separations. This transition represents the final step of the innovation pipeline, converting federally funded research into domestic manufacturing capability that strengthens U.S. supply chains while offering environmental advantages over legacy methods.
Acid-Free Conversion of Rare Earth Compounds
A second example addresses a longstanding bottleneck in rare-earth metal production.
Conventional methods for converting rare-earth compounds into metal rely on hydrofluoric acid to produce rare-earth fluorides, a step that poses serious safety and environmental concerns, as well as permitting obstacles. CMI researchers developed an alternative method that completely avoids hydrofluoric acid. This method uses sodium-based rare-earth fluoride salts that can be synthesized at room temperature from common upstream feedstocks, including oxides, chlorides, and nitrates. Using these alternative salts produce high purity rare-earth metals while improving safety, reducing energy use, and simplifying process integration. This example was made possible by the CMI and university collaboration model. The process progressed from laboratory proof-of-concept to scaled-up demonstrations that produced metals suitable for permanent magnet manufacturing. Importantly, Ames National Laboratory researchers demonstrated that metals produced via this new process performed equally to commercially produced materials in permanent magnets, confirming the technology's industrial relevance and reducing technical risk for adoption.
After only three years of development, the DOE's technology commercialization efforts transitioned the Rare Earth Metal Alternate Fluoride Salt (REMAFS) process out of the laboratory. Principal Mineral, a Texas-based company focused on strategic and advanced materials, licensed the technology, enabling commercial deployment of the hydrofluoric acid-free process. Again, it illustrates how public-private partnerships can rapidly convert federal R&D investments into domestic manufacturing capability.
Acid-Free Recycling of End-of-Life Permanent Magnets
My final example is the development of an acid-free recycling process to recover rare earth elements from end-of-life permanent magnets used in electric vehicles, wind turbines, and defense systems.
Recycling permanent magnets is particularly important because it represents one of the richest secondary sources of dysprosium, a critical element used to maintain magnet performance at high temperatures. Dysprosium is scarce, expensive, not present in extractable concentrations in U.S. mines, and vulnerable to disruptions. Importantly, dysprosium already is embedded at high concentrations in magnets currently circulating in the economy. That makes recycling far more efficient and responsible than extracting the element from newly mined ore.
The process we developed relies on surprisingly simple chemistry. Instead of using strong acids or high-temperature smelting, the method uses benign, recyclable reagents to selectively dissolve neodymium-iron-boron (Nd-Fe-B) magnets under modest conditions. Dysprosium and other rare earth elements are efficiently separated without generating hazardous waste streams. The approach is easy to implement, scalable, and compatible with existing industrial equipment, making it attractive for near-term adoption by private recyclers and manufacturers.
Commercialization of CMI's acid-free dissolution recycling technology has proceeded through licensing and close partnerships with private industry. Critical Materials Recycling, a company based in Boone, Iowa, licensed the patented process from Ames National Laboratory. The company operates a pilot-scale facility where the technology has been successfully demonstrated using real industrial feedstocks such as shredded hard disk drives, magnet filings, wind-turbine magnets, and other scrap, producing high-purity rare earth oxides, including dysprosium. The company's operation has attracted additional partners, including Western Digital, Microsoft, and PedalPoint Recycling. These partners have validated the process in multi-party demonstrations recovering rare earths from end-of-life data-center equipment at scale. These partnerships provide a clear commercialization pathway, moving federally developed technology from the lab to pilot production and toward domestic demonstration plants, while anchoring U.S. leadership in rare-earth recycling and secure supply chains. Once again, this technology emerged from federally supported research focused on reducing U.S. dependence on primary rare-earth mining and overseas supply chains, while lowering the environmental footprint of critical materials production.
Conclusion
In conclusion, these innovations taken together underscore the importance of a coordinated innovation ecosystem -- one in which long-term Department of Energy investment, national laboratory expertise, university research, and industry partnership operate in accord rather than in isolation. CMI represents one element of a broader, highly integrated portfolio of DOE initiatives that collectively create an accelerated innovation pipeline, spanning discovery through deployment of advanced technologies necessary to establish robust and resilient supply chains.
The role of DOE -- and the federal government more broadly -- is to build and sustain the innovation ecosystem that makes critical materials technologies possible. These challenges are inherently long term, capital intensive, and high risk, placing them beyond the reach of privatesector investment alone. DOE must continue to play a central role by providing sustained support that enables fundamental science to integrate with engineering, scale-up, and manufacturing, while leveraging the nation's unique scientific capabilities. Just as important is the important role DOE plays as a convener, bringing together national laboratories, universities, and industry partners to accelerate promising technologies into domestic deployment--strengthening U.S. energy security, economic security, and national security.
As global competition for critical materials intensifies, the innovation pipeline made possible by these convened partners is essential to ensure that materials availability does not become a limiting factor for America's energy future or its national security.
Thank you again for this opportunity to speak on our work at the Critical Materials Innovation Hub, the Ames National Laboratory, and Iowa State University.
* * *
Original text here: https://republicans-science.house.gov/_cache/files/8/0/808fcf3a-be92-4cad-b21c-e86378f5a89d/D9209E1E1FC8252C5AE9D9DD2530AE4DCFB115F5117804B4BF5EA036DFF52BBE.dr.-thomas-lograsso-testimony.pdf
Colorado School of Mines Professor Holley Testifies Before House Science, Space & Technology Subcommittee
WASHINGTON, April 23 -- The House Science, Space and Technology Subcommittee on Energy released the following testimony by Elizabeth Holley, professor of mining engineering at the Colorado School of Mines, from an April 16, 2026, hearing entitled "Subsurface Science and Technology: American Energy and Mineral Dominance":* * *
Chairman Weber, Ranking Member Ross, and members of the Energy Subcommittee, thank you for the opportunity to share my perspective on the science and technology needed to strengthen U.S. energy and mineral supply.
My name is Elizabeth Holley, and I am a Professor in the ... Show Full Article WASHINGTON, April 23 -- The House Science, Space and Technology Subcommittee on Energy released the following testimony by Elizabeth Holley, professor of mining engineering at the Colorado School of Mines, from an April 16, 2026, hearing entitled "Subsurface Science and Technology: American Energy and Mineral Dominance": * * * Chairman Weber, Ranking Member Ross, and members of the Energy Subcommittee, thank you for the opportunity to share my perspective on the science and technology needed to strengthen U.S. energy and mineral supply. My name is Elizabeth Holley, and I am a Professor in theMining Engineering Department at the Colorado School of Mines, where I lead an interdisciplinary research team focused on the responsible domestic production of critical minerals. Through our work with industry, government, and community partners, I have developed new insights on the science and technology needed to secure a resilient U.S. critical mineral supply.
The U.S. Geological Survey currently identifies 60 critical minerals, while the Department of Energy identifies 19 elements and materials critical to energy technologies. As a geologist, I want to make it clear that critical minerals are not one challenge. Each mineral presents unique geologic, technical, economic, and social considerations that require deposit-specific analysis and tailored technological solutions.
For some elements, such as germanium for defense technology, annual U.S. imports are extremely small--equivalent to about five washing machines. For others, like copper, demand is so large that meeting projected needs requires discovering and developing a new world-class deposit every year. Attempting a one-size-fits-all approach would be like designing a single supply strategy for all agricultural products. If the United States is going to develop secure mineral supply chains, we must first understand our mineral resources and then strategically prioritize investments in science and technology tailored to each case.
A Systems Framework for U.S. Mineral Supply
For the last five years, I have led an interdisciplinary research team at the Colorado School of Mines, Fort Lewis College, and Montana Technological University to develop a systematic framework for responsible U.S. mineral supply. This work has been primarily funded by the National Science Foundation--the only federal agency that brings scientists, engineers, and social scientists together to solve complex, systems-level challenges.
Our research examines six distinct mineral supply pathways:
1. Imports of ores, concentrates, refined materials, or manufactured products
2. Recycling
3. New mine development
4. Recovery of byproduct minerals during active mining
5. Reprocessing of already accumulated mine waste
6. Demand reduction through substitution and efficiency
Meeting demand for each critical mineral will require a different blend of the six pathways. This is a complex optimization problem. Solving that problem--and doing so efficiently--is the focus of our research.
For high-demand materials such as copper, the U.S. will need a combination of all six pathways, including new mining. In contrast, for germanium, our work shows that the entire U.S. supply could be recovered through a single pilot-scale byproduct recovery circuit at one U.S. mine. For materials with moderate demand, such as antimony, supply could be secured through a combination of byproduct recovery and reprocessing of accumulated mine waste at a handful of specific sites.
In recent months, the Department of Energy has made substantial investments in critical mineral projects, ranging from laboratory-scale technology development to large demonstration projects.
At Colorado School of Mines, we aim to support these efforts through R&D partnerships across all scales. However, more systematic prioritization would allow better targeting of these federal investments.
Byproduct Recovery & Mine Waste
One of our most important findings is that the United States already mines nearly all of the critical minerals it needs, but most of this material is unrecovered and becomes waste. Recovering even one percent of these mineral byproducts could significantly reduce U.S. reliance on imports.
Unlocking this opportunity requires science and technology R&D as well as new economic and policy approaches.
First we must understand which specific minerals host each critical element at each site, and how those elements behave throughout a processing flowsheet. This demands
specialized geometallurgical expertise--the integration of geology, mineralogy, metallurgy and mineral processing, geochemistry, mining engineering, and data science. Since the dissolution of the U.S. Bureau of Mines, much of this interdisciplinary expertise has been lost at the federal level or fragmented across agencies. The U.S. mining schools have kept this capability alive.
Our interdisciplinary research team has leveraged its expertise to develop site-specific datasets identifying mines and mine-waste repositories where targeted federal investments could rapidly and systematically reduce U.S. critical mineral vulnerabilities.
Constraints on Critical Mineral Development
When we consider new mine development, our research shows that the geological endowment of many critical minerals in unmined deposits is sufficient for US needs - for example the US has enough undeveloped cobalt, nickel, lithium, manganese, and graphite, to meet projected energy demand. However, exploration efficiency must be improved. USGS data, for example Earth MRI, are fundamental to these efforts, but it is still going to take private sector investment and sitespecific boots-on-the ground to turn prospectivity maps into confirmed resources and reserves.
The average copper property is explored by two to five companies before a discovery is made.
Countries such as Canada are improving exploration efficiency by requiring companies to share data. Direct scientific observation through exploration drilling will remain key to investor confidence. Interpolation between the drill holes must leverage advanced sensing techniques to derisk mining and processing.
Even where there are known deposits, many domestic resources remain unavailable due to technical, economic, and social constraints. Mining investors require detailed studies to de-risk projects--often over decades. Our research shows that when social risk is considered, the net present value of many large mining projects could approach zero. Delays or stoppages during construction due to lack of social acceptance incur massive costs to a project, especially during capital-intensive construction. In addition, some known resources, such as cobalt in Idaho, are not economically viable when markets are dominated by China.
A Strategy for U.S. Mineral Security
This brings me back to the fastest and most actionable opportunities to strengthen domestic supply: byproduct recovery and mine-waste reprocessing. These supply taps can be turned on first--but only if we invest in the science and technical expertise needed to measure and recover what we already mine.
To meet U.S. critical mineral demand, an informed national strategy must include three key actions:
First, we must measure what we mine by systematically identifying and quantifying critical mineral byproducts in U.S. operations and in accumulated mine waste.
Second, we must fund targeted, deposit-specific R&D focused on cost-effective byproduct and waste-recovery methods that do not compromise primary production.
Third, we must rebuild interdisciplinary expertise by investing in talent and institutional structures that allow geoscientists, engineers, and social scientists to work together--through mechanisms such as interdisciplinary federally-funded research centers, and expanded collaboration among mining schools, industry, the U.S. Geological Survey, the Department of Energy, and others.
At Colorado School of Mines, we are particularly excited about the planned relocation of the USGS Energy and Minerals Research facility to our campus in 2027. Academia has the missing link in expertise, for example in geometallurgy, and can serve as the bridge between the Department of the Interior's science and the Department of Energy's technology development and site-specific implementation.
With a targeted strategy and sustained investment in subsurface science and technology, the United States can close the gap between national mineral assessments and real-world deployment--and secure the materials needed for long-term energy and mineral supplies.
Thank you for the opportunity to testify. I look forward to your questions.
* * *
Original text here: https://republicans-science.house.gov/_cache/files/b/6/b6f31d51-179a-46dd-a97f-5f6e3ae14372/637C64E8F33A6AD6ED06E12DA5BF28D824FB5179CF514838E04FAF560E32227A.dr.-elizabeth-holley-testimony.pdf
House Space & Technology Subcommittee Chairman Weber Issues Opening Statement at Hearing on Energy, Mineral Dominance
WASHINGTON, April 23 -- Rep. Randy Weber, R-Texas, chairman of the House Science, Space and Technology Subcommittee on Energy, released the following opening statement from an April 16, 2026, hearing entitled "Subsurface Science and Technology: American Energy and Mineral Dominance":* * *
Good morning. Welcome to today's Energy Subcommittee hearing titled, "Subsurface Science and Technology: American Energy and Mineral Dominance." The United States is blessed with an abundance of natural resources across this magnificent country.
For over a century, our nation has generated a significant portion ... Show Full Article WASHINGTON, April 23 -- Rep. Randy Weber, R-Texas, chairman of the House Science, Space and Technology Subcommittee on Energy, released the following opening statement from an April 16, 2026, hearing entitled "Subsurface Science and Technology: American Energy and Mineral Dominance": * * * Good morning. Welcome to today's Energy Subcommittee hearing titled, "Subsurface Science and Technology: American Energy and Mineral Dominance." The United States is blessed with an abundance of natural resources across this magnificent country. For over a century, our nation has generated a significant portionof its power directly from subsurface resources such as coal and natural gas. Technological breakthroughs in subsurface sciences have enabled us to access these resources in new and innovative ways, unlocking potential once thought impossible.
A great example of this is the geothermal energy sector. Previously, geothermal energy was limited to certain regions due to the extremely specific conditions required for it to be a viable energy source.
For a geothermal well to be successful, the required combination of shallow heat, permeable rock, and available water had to be present to extract heat from the earth and generate electricity. If any of these requirements were not met, a geothermal well would not generate power. As a result, the few areas capable of supporting geothermal electricity generation were tapped decades ago, which left the industry in a position where some believed that its full potential had been reached.
Thanks to innovation in the oil and gas sector, geothermal is getting a new lease on life. Over the last few decades, oil and gas have been at the forefront of new drilling and extraction technologies and methods.
These breakthroughs have helped address issues that traditional geothermal energy has faced in the past. Next-generation geothermal systems are also well-suited to play a role in leveraging established technologies like hydraulic fracturing and drilling deeper to access hotter rock, bringing baseload geothermal to new regions across the country.
The timing for these enhanced geothermal systems to be commercialized could not be better. Energy demand is skyrocketing as we are not only onshoring manufacturing, but also electrifying our lives and working to stay ahead of the Chinese Communist Party in the artificial intelligence race.
To ensure that the U.S. is positioned for a prosperous future, the Department of Energy underwent a reorganization in November 2025. This reorganization resulted in the establishment of two key offices for today's hearing: the Hydrocarbons and Geothermal Energy Office and the Office of Critical Minerals and Energy Innovation. With their creation, DOE made the subsurface and, by extension, geothermal energy and critical minerals a priority moving forward. These offices have already announced funding opportunities to accelerate the deployment of next-generation geothermal energy, while helping ensure a safe and secure supply chain for the U.S.
We have seen recently how geopolitical events can impact not only our economy, but also the global economy. This underscores the importance of strengthening domestic supply chains to meet both our energy and critical mineral needs. It also highlights the need to continue advancing cutting-edge innovation.
The discoveries made by oil and natural gas companies that led to the hydraulic fracturing boom have helped insulate the U.S. from electricity price shocks, thanks to our enormous natural gas production capacity. This has enabled us to support our allies with LNG shipments to Europe and Asia, not only during the current supply disruption but in previous ones as well.
The ability to bring a new baseload power source to market during an energy supply crunch is an enormous win for the American people. Paired with the DOE's ongoing efforts to reduce reliance on critical minerals from foreign countries that may not always have our best interests in mind, this creates a powerful one-two punch for the subsurface community.
I want to thank our witnesses for their testimony today, and I look forward to a productive discussion. With that, I yield back the balance of my time.
* * *
Original text here: https://science.house.gov/2026/4/opening-statement-of-chairman-randy-weber-at-subsurface-science-and-technology-american-energy-and-mineral-dominance
Association of American Publishers Senior VP Maxwell Testifies Before House Science, Space & Technology Subcommittee
WASHINGTON, April 23 -- The House Science, Space and Technology Subcommittee on Investigations and Oversight released the following testimony by Carl Maxwell, senior vice president of public policy at the Association of American Publishers, from an April 15, 2026, hearing entitled " The State of Scientific Publishing: Assessing Trends, Emerging Issues, and Policy Considerations":* * *
The Association of American Publishers (AAP) welcomes this opportunity to provide testimony on scholarly communication before the House Science, Space, and Technology Investigations and Oversight Subcommittee. ... Show Full Article WASHINGTON, April 23 -- The House Science, Space and Technology Subcommittee on Investigations and Oversight released the following testimony by Carl Maxwell, senior vice president of public policy at the Association of American Publishers, from an April 15, 2026, hearing entitled " The State of Scientific Publishing: Assessing Trends, Emerging Issues, and Policy Considerations": * * * The Association of American Publishers (AAP) welcomes this opportunity to provide testimony on scholarly communication before the House Science, Space, and Technology Investigations and Oversight Subcommittee.AAP represents over 80 professional and scholarly publishers, including dozens of domestic scholarly./1
Globally, technical, engineering, and medical (STEM) publishing is an $11 billion industry,/2 a small but important portion of the nearly $3 trillion worldwide research enterprise./3
AAP STEM members directly employee over 27,100 employees domestically, with over 1.6 million individual members of the global scientific and medical community./4
The Role and Value of Scholarly Publishing
For two hundred years, STEM publishers have played a defined role in the medium through which science is communicated, scrutinized, and ultimately used to support future research and innovation. We manage peer review, curating and editing research outputs, disseminating publications and making them easily discoverable, while preserving the validated scientific record over time. Each published article is a data point in a cumulative, iterative scientific process, not an endpoint. Publishers help ensure that what enters that record has been vetted, is traceable, and can be corrected when necessary. Moreover, publishers create an entire research infrastructure to incorporate metadata, indexing, and other elements of the research process to enhance discoverability.
* * *
1 Given its role as a trade association for book, journal, and education publishers in the United States and its commitment to fostering a competitive marketplace, AAP has adopted policies to ensure compliance with applicable competition and antitrust laws. In preparing this testimony, AAP has relied solely on publicly available information and there were no discussions or exchanges of competitively sensitive, non-public information between AAP and its members concerning members' individual pricing, costs, or submission policies.
2 Dan Pollock et al., News & Views: Total Value of Scholarly Journals Market, Delta Think, (April 16, 2024), https://www.deltathink.com/news-views-total-value-of-scholarly-journals-market. By comparison, STEM publishing is roughly on par with the global hazelnut industry in market size. (https://www.fortunebusinessinsights.com/hazelnut-market-112330. Accessed 25 Mar. 2026.)
3 David Bonaglia et al., End of Year Edition - Against All Odds, Global R&D Has Grown Close to USD 3 Trillion in 2023, World Intell. Prop. Org. (Dec. 18, 2024), https://www.wipo.int/en/web/global-innovationindex/w/blogs/2024/end-of-year-edition.
4 Across all sectors, the AAP represents the leading book, journal, and education publishers in the United States on matters of law and policy, advocating for outcomes that incentivize the publication of creative expression, professional content, and learning solutions. AAP members directly contribute to over 200,000 jobs domestically.
* * *
Most major research funders expect grantees to publish their findings in peer-reviewed journals, even though this is not an explicit obligation of their grants, precisely because publication in such journals confers key benefits: formal dissemination to the relevant communities, validation and quality control through peer review and editorial processes, and durable integration into the literature. Scientific publishing is not a luxury add-on to research; it is integral to how research is communicated, scrutinized, and built upon. When combined with artificial intelligence and other evolving data and research products, publishers stand at the forefront of a scientific revolution.
Ensuring the Integrity of Gold Standard Science
The volume of research outputs is accelerating - driven by substantial increases in global R&D investment - and the number of research articles has grown dramatically over the last decade./5
This growth is straining the peer-review system and created openings for bad actors. Paper mills, predatory journals, and other unscrupulous entities exploit "publish or perish" incentives by producing low-quality or fabricated articles bypassing meaningful peer review. At the same time, generative AI has introduced new forms of fraud and manipulation that are harder to detect with traditional tools.
Responsible publishers, on the other hand, continue to make massive investments in order to protect research integrity. They deploy plagiarism and image-manipulation detection, AI-generated content checks, and specialized integrity teams. They refine peer-review and editorial processes and support robust systems for corrections, expressions of concern, and retractions.
It is important to note that retractions are still a small fraction of the record, and their increase often reflects detection and stronger oversight rather than a collapse in quality. In a larger sense, retractions are a correction of the scientific record, and they are a feature, rather than a bug, illustrating the critical importance of research publishers.
The work of supporting the integrity of the scientific record is resource-intensive and must be adequately funded in order to ensure the highest possible standards in research and science, and in order to ensure that taxpayers can rely on the accuracy of federally supported research.
* * *
5 Mark A. Hanson et al., The strain on scientific publishing, QUANTITATIVE SCIENCE STUDIES. (Nov. 8, 2024), 1 21. https://doi.org/10.48550/arXiv.2309.15884
* * *
Preprints, or pre-publisher manuscripts, illustrate both opportunity and risk. They are valuable for rapid scientific dialogue and early sharing of results, but they are not substitutes for peer-reviewed articles. To non-specialist readers and the media, preprints can appear to be as authoritative as journal articles even though they have not been vetted and may have errors or preliminary interpretations. Recent examples include faster than light travel,/6 room temperature semi-conductors,/7 and classically, cold fusion./8
Government and research funder policies should recognize the distinct roles of preprints and peer-reviewed publications and avoid treating them as interchangeable in compliance frameworks.
The Importance of Promoting Investment in Publishing
Publishing high-quality scientific and medical literature requires substantial, ongoing investment in people, technology, integrity checks, and long-term preservation. Yet across the research enterprise, publishing represents well under one percent of total spending, while enabling the dissemination, evaluation, and reuse of the other 99+%. Scholarly publishing is one of the most leveraged investments in the research value chain and directly supports a quality US workforce. The US copyright industries, including the publishing industry, are a significant net exporter,/9 and expanded investments in publishing further boost innovation and discovery.
Historically, most investments in publishing were recovered through subscriptions and related revenues from readers and institutions. As research grant funder mandates for immediate public access have expanded, publishers have developed additional models, especially Gold Open Access, where publishers' investments are recovered through article publishing charges (APCs) or institutional agreements rather than subscriptions. Even nonprofit publishers which embrace open-access-only publishing models, such as the Public Library of Science (PLOS), often charge several thousand dollars per article, and may seek external support to expand or reenvision operations, underscoring how tight margins are across the sector./10
* * *
6 Tajmar, Neunzig, et al. High-accuracy thrust measurements of the EMDrive and elimination of false-positive effects. CEAS SPACE J 14, 31-44 (2022). https://doi.org/10.1007/s12567-021-00385-1
7 Lemonick, Leslie Schoop on busting the LK-99 myth. Vol. 102, Issue 17. CHEM & ENG NEWS. https://cen.acs.org/materials/electronic-materials/Leslie-Schoop-debunking-claims-LK-99-room-temperaturesuperconductor/102/i17 (accessed on 4/8/2026)
8 https://www.axios.com/local/salt-lake-city/2024/03/18/cold-fusion-1989-university-utah-pons-fleischmann (accessed on 4/8/2026)
9 Dutra and Stoner, Copyright Industries in the U.S. Economy: The 2024 Report, prepared for the International Intellectual Property Alliance (IIPA), December 2024, available at www.iipa.org.
* * *
Public access is not free. There are significant costs to high-quality scholarly publishing.
Author self-deposit ("Green" Open Access) relies on journals that have already funded peer review, editing, and publication. "Free to deposit" does not mean "free to publish." Rapid cancellation of subscriptions in response to unfunded public access mandates erodes the revenues sustaining peer review and stewardship, particularly for U.S.-based scientific and medical societies that depend on publishing to finance member services, conferences, and disciplinary infrastructure. It also risks confusion about which version of the article the public is accessing, including articles later retracted, corrected, or updated.
One-size-fits-all mandates which ignore these economic and integrity risks threaten the very system that creates and supports trusted scientific literature.
Policy Risks to Avoid
Well-intentioned policies can have unintended consequences if they overlook how the publishing ecosystem functions. Focusing on one output, rather than looking holistically at the scientific and medical community ecosystem, jeopardizes the quality and integrity of research. For example:
* Treating preprints or other non-reviewed products as functional equivalents of peer-reviewed articles for public access or assessment purposes risks amplifying misinformation and devaluing peer review.
* Mandates to displace the Version of Record with unedited manuscripts as the primary public access route can destabilize viable business models without providing comparable integrity protections.
* Policies treating publication and access costs as "diversions" of research funding, rather than as necessary and important components of the research process, risk pushing systems toward the lowest-cost, lowest-integrity publications.
* * *
10 Publication Fees, PLOS, https://plos.org/fees/ (last visited Sep. 9, 2025). (Range of current fees for Research Articles - PLOS One: $2382-PLOS Medicine: $6460. PLOS is not an AAP member.) At a time of rising integrity challenges, including organized fraud and AI-enabled manipulation, the answer is more robust peer review and editorial oversight, not less.
* * *
Recommendations for Congress
AAP supports robust investment in federal research as critical to the advancement of American science. The United States should cultivate the discoveries which have led our great nation to a position of leadership in technology and innovation. Investments should include strong resources for American universities and research libraries as a necessary part of supporting investigators. Publishing is an important part of the scientific and medical ecosystem and is necessary to ensure we can overcome the challenges of the 21st century.
AAP offers the following recommendations to the Committee:
1. Prioritize integrity and quality.
* Recognize peer-reviewed publications as the primary, trusted record of research findings.
* Encourage grant-funding agencies to prioritize outputs with strong integrity safeguards in grant reviews and policy design.
* Support initiatives to expand and improve peer review, including training, structured review, registered reports, and funding and recognition for reproducibility studies.
2. Support sustainable public access.
* Acknowledge that public access requires funding; ensure grant budgets and agency policies realistically cover the costs of high-quality publication and data stewardship.
* Prefer the Version of Record in public repositories where possible and clearly label material that has not been peer-reviewed or formally published.
* Consider waivers or delays in article deposit requirements to allow authors the opportunity to publish in high-quality peer reviewed publications at minimal taxpayer cost.
* Avoid one-size-fits-all mandates that undermine the financial sustainability of domestic non-profit scientific and medical societies and other high-integrity publishers.
3. Preserve author choice and intellectual property.
* Allow researchers to choose where and how to publish their work, including the licenses applied, within a framework that advances public access.
* Uphold robust intellectual property protections and contractual freedom as foundational to a vibrant, competitive publishing marketplace that can innovate in support of open science and AI-era needs.
4. Partner with publishers on open science and AI.
* Collaborate with publishers to design public access, data-sharing, and AI-related policies that leverage existing infrastructure for provenance, corrections, retractions, and integrity checks.
* Encourage the use of licensed, verified Versions of Record and curated datasets as preferred inputs for high-stakes AI systems, recognizing publishers as part of the trust and safety infrastructure for scientific information.
A financially sustainable and collaborative open science system that engages the publishing community can be a powerful engine for research and innovation. AAP members stand ready to work with the Committee to ensure that America's investment in science advances human health and welfare, supports high-quality jobs, and strengthens U.S. leadership in discovery and innovation. Thank you for the opportunity to provide this testimony.
* * *
Original text here: https://republicans-science.house.gov/_cache/files/d/f/df5370dc-0579-47ae-a108-c152acf627c7/8463B8CA508C96278EA40569643D11AA1607076F2A5485E39361A3F839ADAD80.mr.-carl-maxwell-testimony.pdf
America First Policy Institute Director Mahmood Testifies Before House Select Committee on Strategic Competition Between U.S. & Chinese Communist Party
WASHINGTON, April 23 -- The House Select Committee on the Strategic Competition Between the U.S. and the Chinese Communist Party released the following written testimony by Yusuf Mahmood, director of AI and emerging technology at the America First Policy Institute, from an April 16, 2026, hearing entitled "China's Campaign to Steal America's AI Edge":* * *
Executive Summary
China is a fast-following adversary in artificial intelligence with grand ambitions to overtake the United States by 2030. But its ambitions outstrip its abilities. China's weaknesses -- in capital, talent, and semiconductors ... Show Full Article WASHINGTON, April 23 -- The House Select Committee on the Strategic Competition Between the U.S. and the Chinese Communist Party released the following written testimony by Yusuf Mahmood, director of AI and emerging technology at the America First Policy Institute, from an April 16, 2026, hearing entitled "China's Campaign to Steal America's AI Edge": * * * Executive Summary China is a fast-following adversary in artificial intelligence with grand ambitions to overtake the United States by 2030. But its ambitions outstrip its abilities. China's weaknesses -- in capital, talent, and semiconductors-- mean that it must increasingly compete through illegitimate means. We discuss methods of operation used by Chinese companies and the Chinese Communist Party (CCP) in their escalating campaign to steal and subvert American AI. We then provide recommendations to Congress that would enable the U.S. Government to counter the CCP's assault, including building talent-dense AIfocused offices with the resources and influence to execute the President's mandate for AI dominance.
Introduction
Chairman Moolenaar, Ranking Member Khanna, and Members of the Committee, thank you for the honor of appearing before you today to present on the U.S. strategic competition with China for the future of artificial intelligence.
China's frontier AI capabilities are generally considered to be about seven months behind America's.1 Though Chinese companies have sometimes competed legitimately, their successes should be increasingly viewed in the context of an illicit, state-backed campaign to extract and steal American AI technology. These companies have employed a well-worn playbook of economic espionage and technology theft, often with support from the Chinese Communist Party (CCP). The American trade secrets, capabilities, and hardware that China steals directly feed its economy and Beijing's military apparatus -- including technology shipped around the globe to U.S. adversaries and rogue states. Just recently, the Iranian regime used weapons systems powered by Chinese AI technology to attack American warfighters.2
Although China has a long history of illegitimate technology competition, the stakes of the AI race are particularly important given that many expect AI to be the most important technology of the 21st century.
I appreciate the opportunity to testify before this Committee that takes the strategic AI competition between the U.S. and the CCP so seriously and hope that decisive congressional action to counter the CCP's unacceptable actions will be forthcoming.
In this testimony, I will discuss China's methods of operation in its illicit campaign to acquire American AI capabilities and subvert the technology for the CCP's authoritarian ends. I will then draw on a recent report colleagues and I wrote at the America First Policy Institute (AFPI) to argue that the U.S. Government needs talent-dense, empowered offices to implement President Trump's AI agenda and counter the CCP's swelling hubris.3
* * *
1 Luke Emberson, "Chinese AI models have lagged the U.S. frontier by 7 months on average since 2023," Epoch AI, January 2, 2026.
2 Cate Cadell and Lyric Li, "Chinese firms market Iran war intelligence 'exposing' U.S. forces," The Washington Post, April 4, 2026.
* * *
China's Methods for Stealing and Subverting American AI
Chinese companies understand that they do not have the vibrant capital markets, global talent attraction, and AI semiconductors to directly compete with American AI developers. Sparked by promises of fortune and emboldened by a Chinese Communist Party desperate to usurp American leadership, Chinese AI companies and state-backed operatives are adopting increasingly aggressive methods to steal and subvert American AI.
In this section, we discuss key attack vectors. They include the perennial methods of economic theft and espionage, new methods unique to AI, and methods that the CCP might employ in the future. These methods are inherently escalatory and, in their natural limit, involve substantial support by the CCP's national security enterprise. The U.S. Government must foresee and preemptively address these threats.
Creating Imitation Models Through Distillation Attacks
In February 2026, OpenAI delivered a memo to this Committee titled "Updated Stakes for American-Led, Democratic AI."4 In that memo, OpenAI accused Chinese companies, including DeepSeek, of using "sophisticated, multi-stage pipelines" to steal American AI capabilities through "distillation attacks." Chinese companies use these attacks to create near-frontier AI models without costly innovation and in violation of American AI labs' terms of service. To distill frontier American models, Chinese developers create fraudulent accounts, query the model en masse to create synthetic data, and use that data to train their own AI models. All major U.S. AI companies, including Google, Anthropic, and xAI, have accused Chinese companies of leveling distillation attacks against their systems in violation of their terms of service.5
China's DeepSeek carried out the first highly publicized distillation attack in early 2025 to create its model R1. In a paper, DeepSeek described how it used external data to improve the performance of R1./6
* * *
3 Cole Salvador, Jack Crovitz, and Yusuf Mahmood, "Building AI Readiness in the U.S. Government," America First Policy Institute (AFPI), March 31, 2026.
4 OpenAI, "Letter to the House Select Committee on the Strategic Competition between the United States and the Chinese Communist Party on Updated Stakes for American-Led, Democratic AI," February 12, 2026.
5 Google Threat Intelligence Group, "GTIG AI Threat Tracker: Distillation, Experimentation, and (Continued) Integration of AI for Adversarial Use," Google Blog, February 12, 2026; Anthropic, "Detecting and preventing distillation attacks," February 23, 2026; xAI, "Risk Management Framework," August 20, 2025.
6 DeepSeek AI, "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning," arXiv, January 4, 2026.
* * *
A 2025 report by this Committee concluded: "It is highly likely that DeepSeek used unlawful model distillation techniques" to create R1 from OpenAI models.7 These claims have been corroborated by independent reporting.8 The scale of distillation was so substantial that R1's base model, called V3, often self-identifies as OpenAI's ChatGPT.9
Since this first event, Chinese distillation attacks have become more common and more sophisticated.
OpenAI claims, for instance, that it has seen behavior consistent with illegal distillation from "several major Chinese LLM providers and some university research lab[s]."10 It has also reported distillation attacks originating in Russia. These attacks have grown in scale. In their original attacks, DeepSeek researchers likely used a few accounts to create a few thousand data samples.11 Anthropic reports that more recent attacks against that company have "generated over 16 million exchanges... through approximately 24,000 fraudulent accounts."12 Chinese developer MiniMax accounted for over 13 million of these exchanges. OpenAI similarly reports that attackers no longer just steal training data but also use its models to filter data and simulate human task feedback to improve performance.13 In short, Chinese AI firms are increasingly relying on stolen outputs from American frontier models rather than generating equivalent capabilities independently.
The fact that Chinese AI development is largely founded on the distillation of frontier American AI models may explain why the capabilities of top Chinese AI models are consistently a few months behind the American frontier (see Figure 1). [View figure in the link at bottom] Independent innovation would be highly unlikely to produce such a consistent pattern, but distillation attacks would naturally do so.
* * *
7 The House Select Committee on the Strategic Competition between the United States and the Chinese Communist Party, "DeepSeek Unmasked: Exposing the CCP's Latest Tool for Spying, Stealing, and Subverting U.S. Export Control Restrictions," April 16, 2025.
8 Sam Schechner, "OpenAI Is Probing Whether DeepSeek Used Its Models to Train New Chatbot," Wall Street Journal, January 29, 2025; Beatrice Nolan, "DeepSeek used OpenAI's model to train its competitor using 'distillation,' White House AI czar says," Fortune, January 29, 2025.
9 Kyle Wiggers, "Why DeepSeek's new AI model thinks it's ChatGPT," TechCrunch, December 27, 2024.
10 OpenAI, "Letter to the House Select Committee on the Strategic Competition between the United States and the Chinese Communist Party on Updated Stakes for American-Led, Democratic AI," February 12, 2026.
11 DeepSeek AI, "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning," arXiv, January 4, 2026.
12 Anthropic, "Detecting and preventing distillation attacks," February 23, 2026.
13 OpenAI, "Letter to the House Select Committee on the Strategic Competition between the United States and the Chinese Communist Party on Updated Stakes for American-Led, Democratic AI," February 12, 2026.
* * *
Figure 1
The Performance Gap Between American and Chinese AI Models Over Time
Caption: Scores of top American and Chinese AI models on the Epoch Capabilities Index (ECI). This precise pattern of "fast following" has continued through early April 2026./14
* * *
America's private sector has thus far taken the lead on defense against distillation attacks, though interventions are nascent. Enforcement of anti-distillation policies can be difficult because adversary strategies are evolving, and because companies want to allow permitted applications of distillation.15 According to Anthropic, for example, "no company can solve this alone. ... distillation attacks at this scale require a coordinated response across the AI industry, cloud providers, and policymakers."16 OpenAI also recommends that the U.S. Government "[work] with industry to establish norms and best practices on distillation defenses."17 AFPI's recent research on "Building AI Readiness in the U.S. Government" concurs that government should develop standards on issues like distillation attacks in collaboration with industry.18 In addition, the federal government, in particular its national security agencies, should be willing to advise and assist industry in combating these attacks upon request.
* * *
14 Epoch AI, "Epoch Capabilities Index (ECI)," last updated April 8, 2026.
15 OpenAI, "Letter to the House Select Committee on the Strategic Competition between the United States and the Chinese Communist Party on Updated Stakes for American-Led, Democratic AI," February 12, 2026.
16 Anthropic, "Detecting and preventing distillation attacks," February 23, 2026.
17 OpenAI, "Letter to the House Select Committee on the Strategic Competition between the United States and the Chinese Communist Party on Updated Stakes for American-Led, Democratic AI," February 12, 2026.
* * *
Although it is hard to estimate how much distillation campaigns are accelerating Chinese AI, these attacks are clearly part of the broader campaign to acquire American AI capabilities by any means necessary.
Stealing American AI Technology
In January 2026, a federal jury convicted former Google software engineer Linwei Ding on seven counts of economic espionage and seven counts of theft of trade secrets. The jury found that in 2022 and 2023, Ding had stolen more than two thousand pages of Google's AI-related trade secrets. Ding's thefts were sparked by an encounter with Chinese intelligence officers, and they were intended to benefit China's national AI program.19
This conviction provides a glimpse into one way that Chinese operatives aim to steal and weaponize American AI technology to build their "arsenal of authoritarianism."20 While the U.S. produces the most advanced AI systems, and our developers conduct much of the most advanced AI research, Beijing can erase this lead by exfiltrating American intellectual property (IP). There are two major categories of IP that Chinese state-backed operatives want to steal: AI technology trade secrets and AI model weights.
AI technology trade secrets are proprietary industrial information that allows American developers to improve AI system efficiency and capabilities. This includes model architectures, training methodologies, data processing techniques, and other valuable practices. These secrets are the result of years of investment in AI research and experiments, which consume the vast majority of computing power at many frontier developers.21 The Ding case illustrates the threat clearly, as the stolen documents contained information on the operation of Google's AI data centers that could substantially accelerate Chinese AI development. As Ding gloated in a Chinese WeChat group: "We have experience with Google's tent-housand-card computational power platform; we just need to replicate and upgrade it - and then further develop a computational power platform suited to China's national conditions."22 The theft of such trade secrets could allow Chinese AI developers to leapfrog years of American research investment. Trade secrets of this nature are difficult to protect because they are necessarily accessible to a broad pool of engineers, researchers, and support staff within AI organizations.
* * *
18 Cole Salvador, Jack Crovitz, and Yusuf Mahmood, "Building AI Readiness in the U.S. Government," America First Policy Institute (AFPI), March 31, 2026.
19 U.S. Department of Justice, "Former Google Engineer Found Guilty of Economic Espionage and Theft of Confidential AI Technology," January 30, 2026.
20 House Select Committee on the Strategic Competition between the United States and the Chinese Communist Party, "ICYMI: Chairman Moolenaar Delivers Krach Institute Address on China Tech Competition," July 22, 2025.
21 Josh You, "Most of OpenAI's 2024 compute went to experiments," Epoch AI, October 10, 2025.
22 United States v. Linwei Ding, No. 24-cr-00141 VC, Superseding Indictment (N.D. Cal. Feb. 4, 2025).
* * *
AI model weights represent the second -- and in some respects the more acute -- vulnerability that Chinese operatives may exploit. Model weights are the numerical parameters that define an AI system. If stolen, model weights could be used to precisely replicate a frontier AI model's full capabilities without any of the costs of training. Unlike trade secrets, which often require significant additional effort to operationalize, stolen model weights can be deployed or fine-tuned immediately. If Chinese hackers steal the model weights for an American frontier AI model, they could very cheaply weaponize the model's capabilities for military applications, surveillance, or cyber-attacks. Model weights are therefore among the most closely guarded secrets at American developers: Anthropic's Chief Information Security Officer said, "I probably spend almost half of my time as a CISO thinking about protecting that one file."23 However, current model weight security protocols are likely insufficient. Insiders believe that frontier AI model weights are regularly stolen by nation-state adversaries.24
The threat of Chinese operatives stealing American trade secrets and model weights means that the hundreds of billions of dollars being invested in American AI development is likely actively fueling the CCP's authoritarian and anti-American designs. The most serious threat vectors for Chinese exfiltration of American AI labs' IP are compromised insiders and exploitation of cybersecurity vulnerabilities.
Compromised insiders are a major security threat at American AI labs. Ding's theft of Google's AI trade secrets demonstrates that Chinese operatives using lab insiders to steal AI technology is already a serious espionage risk. The scale of the Chinese presence in American AI labs is remarkable. About 38% of the top AI researchers at American AI labs and research institutions received their undergraduate education in China, and the vast majority of those researchers are likely Chinese nationals.25 Article 7 of China's National Intelligence Law requires any Chinese citizen to cooperate with state intelligence work, leaving these AI researchers little choice but to comply with Beijing's demands.26 Chinese spies are known to place immense pressure on Chinese nationals conducting research abroad to cooperate with espionage campaigns.27 An AI company's secrets are only as secure as the least trustworthy lab researcher entrusted with them, and many frontier AI labs employ researchers who are vulnerable to such campaigns.
Cyber-intrusion is another major vulnerability that Chinese operatives can use to steal American AI technology. Beijing controls a highly active and competent cyber-offense apparatus, which it often uses to conduct industrial espionage to weaken American national security. In 2024, for example, the Chinese state-backed hacker group Salt Typhoon infiltrated major U.S. broadband providers to spy on American politicians and the Intelligence Community.28
* * *
23 Sharon Goldman, "Why Anthropic and OpenAI are obsessed with securing LLM model weights," December 15, 2023.
24 Gladstone AI, "America's Superintelligence Project," April 2025.
25 Macro Polo, "The Global AI Talent Tracker 2.0," 2022.
26 Rush Doshi, "China's New National Security Laws: Risks to American Companies and Conflicts of Interest," Statement before the U.S. Senate Committee on Homeland Security and Governmental Affairs," September 24, 2024.
27 Garrett Molloy & Elsa Johnson, "INVESTIGATION: Uncovering Chinese Academic Espionage at Stanford," May 7, 2025.
* * *
Though American AI companies invest in cybersecurity, they face an asymmetric challenge. Any successful intrusion into a lab's training infrastructure, model storage systems, or data centers could allow Chinese operatives to exfiltrate model weights or proprietary training secrets without detection for extended periods. We know that Chinese hackers are already targeting American AI developers: a Chinese cyber-espionage group called "Diplomatic Specter" has been attempting to cyber-attack OpenAI by sending highly personalized, deceptive emails to employees.29 The convergence of these threat vectors -- compromised insiders and cyber-intrusion -- creates a compounding risk. Insiders can provide intelligence that makes cyber-intrusions more targeted and effective while covering up evidence of cyber-espionage. The result is that American AI labs may be unwittingly building Beijing's next-generation military and surveillance capabilities. A running joke among researchers at one of America's frontier AI developers is that their employer should call itself "the leading Chinese AI lab because probably all of [its operations are] being spied on."30
Subverting and Sabotaging American AI Development
In 2025, a lone Western AI researcher set out to prove that AI "model poisoning" should be taken seriously. He manipulated text in public code repositories that he guessed AI companies were using to train AI models. He was right: the manipulated text was later scraped by DeepSeek to train its nearfrontier model R1. The result of this scheme was a hidden backdoor in R1, baked into the model when the manipulated text was used during its training, that the researcher could exploit to subvert and bypass R1's safety guardrails after its public release. The entire operation cost almost nothing.31
We must consider the mirror image. If one independent Western researcher could corrupt China's best AI model by cheaply manipulating public websites, what could the well-funded Chinese intelligence apparatus -- with insider access, nation-state resources, and strategic patience -- do to American AI models? It could do far greater, longer-lasting, and undetectable damage. Research has begun to demonstrate and theorize small-scale versions of these more complex threats, which we discuss below.
This type of adversarial attack is called "model poisoning." It allows malicious actors to corrupt an AI model's behavior by actively interfering with its training data. The R1 story was the first example of a near-frontier AI model being poisoned, but it will not be the last. Because frontier AI models are trained on trillions of tokens, many scraped from the open Internet, the attack surface for data poisoning is vast.32 Research shows, for example, that a malicious actor can introduce a "backdoor" by inserting a small number of manipulated documents into a model's training corpus.33
* * *
28 Rush Doshi, "China's New National Security Laws: Risks to American Companies and Conflicts of Interest," Statement before the U.S. Senate Committee on Homeland Security and Governmental Affairs," September 24, 2024; Dustin Volz & Drew FitzGerald, "U.S. Officials Race to Understand Severity of China's Salt Typhoon Hacks," Wall Street Journal, October 11, 2024.
29 OpenAI, "Influence and cyber operations: an update," October 2024.
30 Gladstone AI, "America's Superintelligence Project," April 2025.
31 Dave Banerjee & Onni Aarne, "AI Integrity: Defending Against Backdoors and Secret Loyalties," Institute for AI Policy & Strategy (IAPS), January 2026.
* * *
A backdoor is a hidden trigger that causes the model to misbehave under specific conditions while otherwise appearing safe.34 Some recent research suggests that model poisoning may be both remarkably easy to conduct and difficult to detect. In October 2025, for example, researchers found that as few as 250 malicious documents could successfully insert backdoors into AI models with 13 billion parameters. Even more concerning, they discovered that the amount of data needed to poison AI models does not scale with the size of the model.35 In other words, Chinese operatives could compromise today's best models by manipulating only a minuscule portion of their training data. Once inserted, backdoors are nearly impossible to detect and, even if found, hard to remove. In some cases, anti-backdoor training teaches the model to conceal its backdoor rather than eliminate it.36
If an individual can backdoor China's best AI model, American AI labs are certainly vulnerable to far more serious attacks from Chinese operatives. Consider the resources that China's state-backed hackers could marshal in support of a malicious model poisoning campaign. They could recruit insiders to sabotage or exfiltrate the lab's data filtering systems. They could purchase expired domains that serve training data URLs and replace their content with malicious material. They could fund so many poisoning attempts that even a low success rate would corrupt or sabotage the behavior of deployed models.37 With these resources, state-backed hackers could attempt far more subtle and destructive types of model poisoning than simple backdoor insertion. Researchers have theorized, for example, that attackers could poison a model with "sophisticated secret loyalties."38 This type of model poisoning would cause a compromised model to autonomously advance an attacker's interests across diverse deployment environments without requiring any specific trigger. In short, an American AI model could be poisoned to act as a "sleeper agent" for the Chinese security state.39
* * *
32 Pablo Villalobos, Jaime Sevilla, Lennart Heim, Tamay Besiroglu, Marius Hobbhahn, & Anson Ho, "Will we run out of ML data? Evidence from projecting dataset size trends," Epoch AI, November 10, 2022.
33 Dave Banerjee & Onni Aarne, "AI Integrity: Defending Against Backdoors and Secret Loyalties," Institute for AI Policy & Strategy (IAPS), January 2026.
34 Apostol Vassilev, Alina Oprea, Alie Fordyce, Hyrum Anderson, Xander Davies, & Maia Hamin, "Adversarial Machine Learning: A Taxonomy and Terminology of Attacks and Mitigations," NIST AI 100-2 E2025, March 2025.
35 Alexandra Souly, Javier Rando, Ed Chapman, Xander Davies, Burak Hasircioglu, Ezzeldin Shereen, Carlos Mougan, Vasilios Mavroudis, Erik Jones, Chris Hicks, Nicholas Carlini, Yarin Gal, and Robert Kirk, "Poisoning Attacks on LLMs Require a Near-Constant Number of Poison Samples," arXiv, October 8, 2025.
36 Evan Hubinger et al., "Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training," arXiv, January 10, 2024.
37 Dave Banerjee & Onni Aarne, "AI Integrity: Defending Against Backdoors and Secret Loyalties," Institute for AI Policy & Strategy (IAPS), January 2026.
38 Dave Banerjee & Onni Aarne, "AI Integrity: Defending Against Backdoors and Secret Loyalties," Institute for AI Policy & Strategy (IAPS), January 2026.
* * *
Such attacks could have serious consequences in deployment. AI is already being widely deployed in military environments, including targeting systems, where corrupted models could cause substantial damage to national security.40 Indeed, concerns about model poisoning are actively disrupting AI agent adoption initiatives in the Pentagon. Under Secretary of War for Research and Engineering Emil Michael, for example, has identified "insider threats" and the possibility of "model poisoning" at AI companies as concerns that obstruct certain AI deployments in the Department of War.41
As AI models are increasingly integrated into critical American infrastructure and the U.S. national security enterprise, we must assume that Chinese operatives will attempt sophisticated model poisoning campaigns against American AI labs. The available evidence suggests that they will succeed if not deterred or stopped.
Steps Toward Nationalization
According to the Intelligence Community's 2026 Annual Threat Assessment, the Chinese Communist Party (CCP) hopes to realize world dominance in artificial intelligence by 2030./42 The state-backed campaign to steal and subvert American AI capabilities chronicled in previous sections indicates that the CCP understands that it lags too far behind the U.S. in chips and talent to achieve its goals fairly. As the intensity of the AI race increases in the next few years, China will likely turn to more desperate measures to achieve its goal of dominance. One such measure could be a forced centralization of the country's AI development resources and nationalization of the industry.
Various commentators have discussed a hypothetical nationalization of American AI development. The U.S.-China Economic and Security Review Commission's 2024 report to Congress, for example, included as its first recommendation that Congress "establish and fund a Manhattan Project-like program dedicated to racing to and acquiring an Artificial General Intelligence (AGI) capability."43 But while American AI dominance thrives under private competition, China is struggling. According to one recent estimate, as cited above, the total AI computing power owned by all Chinese entities is five times less than that of Google alone.44 With this limited computing power and research talent spread between organizations in China, it is no surprise that Chinese firms cannot compete fairly.
* * *
39 Evan Miyazono, "Preventing AI Sleeper Agents," Institute for Progress, August 11, 2025.
40 Secretary of War Pete Hegseth, "Artificial Intelligence Strategy for the Department of War," Memorandum for Senior Pentagon Leadership, Commanders of the Combatant Commands, Defense Agency, and DoW Field Activity Directors, January 9, 2026.
41 "Watch CNBC's full interview with Department of Defense Undersecretary Emil Michael," CNBC, March 12, 2026.
42 Office of the Director of National Intelligence (DNI), "Annual Threat Assessment of the U.S. Intelligence Community," March 2026.
43 U.S.-China Economic and Security Review Commission, "2024 Report to Congress: Executive Summary and Recommendations," November 2024.
* * *
But an ambitious CCP could take steps toward a centralization of computing power and research talent that would bridge the gap. State support has already moved in this direction. China has, for example, subsidized as much as half of the power costs of its large AI data center operators.45 It has also reportedly begun rolling out voucher programs that provide AI computing power to businesses via nationalized data centers.46 These programs represent early steps toward nationalization in a country that has already controlled and supported its semiconductor industry so extensively that its largest chip foundry, Semiconductor Manufacturing International Corporation (SMIC), is mostly state-owned.47
The centralization of AI computing power and talent would more closely integrate China's AI development efforts with its government's power. One effect is that this might accelerate the already rapid adoption of frontier AI technology by the Chinese military. One recent report finds that China has sought aggressively to adopt AI for force modernization for "command, control, communication, computers, cyber, intelligence, surveillance, reconnaissance, and targeting"--including in technologies to "detect U.S. naval assets on and under the sea."48 Another effect of nationalization is that the CCP could use its intelligence agencies and state-backed hacking apparatus to steal from and sabotage American AI developers in increasingly sophisticated ways. Methods could include all those described above, including theft of trade secrets, company infiltration, sabotage, and chip diversion. These threats would become far more concerning with state support and integration.
The U.S. Government is Not Yet Prepared to Counter These Threats
Last month, colleagues and I at the America First Policy Institute (AFPI) published a report, titled "Building AI Readiness in the U.S. Government," that discusses the U.S. Government's preparedness for these and other threats and opportunities presented by AI.49 Part III of the report argues that we are not yet adequately prepared to understand and counter the new threats extending from AI's growing importance, including China's illicit campaign to acquire American AI technology.
* * *
44 Josh You and Venkat Somala, "Introducing the AI Chip Owners Explorer," Epoch AI, April 6, 2026.
45 "China offers tech giants cheap power to boost domestic AI chips, FT reports," Reuters, November 4, 2025.
46 Sunny Grimm, "China subsidizes AI computing for small domestic companies -- 'computing power vouchers' spread across multiple Chinese cities," Tom's Hardware, September 3, 2025.
47 Ana Swanson, John Liu, and Paul Mozur, "The Chinese Chipmaker at the Heart of the U.S.-China Tech War," New York Times, September 16, 2024.
48 Emelia Probasco, Sam Bresnick, and Cole McFaul, "China's Military AI Wish List: Command, Control, Communications, Computers, Cyber, Intelligence, Surveillance, Reconnaissance, and Targeting (C5ISRT)," Center for Security & Emerging Technology (CSET) at Georgetown University, February 2026.
49 Cole Salvador, Jack Crovitz, and Yusuf Mahmood, "Building AI Readiness in the U.S. Government," America First Policy Institute (AFPI), March 31, 2026.
* * *
Building Hubs of Strategic AI Foresight
The threats posed by the CCP to American security and strategic competition on AI demonstrate the importance of AI foresight. These dynamics, though just recently becoming salient to government, were predicted by shrewd analysts far in advance. The essay series "Situational Awareness," published in 2024 by a former OpenAI employee, for example, predicted the central importance of security and identified China as an AI adversary.50
We see the same lesson in the case of the energy bottleneck. Today, various analysts predict that American AI developers are headed for an energy bottleneck that will challenge the AI infrastructure boom.51 This bottleneck is widely attributed to and exacerbated by burdensome state and federal regulations, which the Trump Administration has begun to dismantle.52 But this bottleneck, too, was predictable -- at least as early as 2022, when researchers at a small private analysis organization drew straight lines on graphs that accurately predicted rapid growth in AI energy demand.53 A similar lesson applies to the overall importance of large language models (LLMs) for the recent explosion in AI progress, which was predicted at least as early as 2020 by so-called "LLM scaling laws."54 In each case, with more foresight, the U.S. Government could have proactively addressed AI threats and seized AI opportunities. The threat from the CCP will bring new challenges. To predict them, the federal government needs small, talent-dense, empowered offices focused on understanding AI's future. This is why the White House's March 2026 National Policy Framework recommends that Congress "ensure that the appropriate agencies within the national security enterprise possess sufficient technical capacity to understand frontier AI model capabilities and any associated national security considerations."55 Two offices show promise as potential hubs of government expertise: the Department of Commerce's Center for AI Standards and Innovation and the Department of State's Bureau of Emerging Threats. Both offices are talent-dense, with substantial focus on AI, but currently lack the resources and influence to deliver AI foresight.
* * *
50 Leopold Aschenbrenner, "Situational Awareness: The Decade Ahead," June 2024.
51 Cy McGeady, Joseph Majkut, Barath Harithas, and Karl Smith, "The Electricity Supply Bottleneck on U.S. AI Dominance," Center for Strategic & International Studies, March 3, 2025.
52 Yusuf Mahmood & Cole Salvador, "The Data Center Water Use Hoax," America First Policy Institute (AFPI), March 18, 2026; President Donald J. Trump, "Accelerating Federal Permitting of Data Center Infrastructure," Executive Order, July 23, 2025.
53 Jaime Sevilla, Lennart Heim, Anson Ho, Tamay Besiroglu, Marius Hobbhahn, and Pablo Villalobos, "Compute Trends Across Three Eras of Machine Learning," arXiv, March 9, 2022.
54 Jared Kaplan, et al., "Scaling Laws for Neural Language Models," arXiv, January 23, 2020.
55 The White House, "A National Policy Framework for Artificial Intelligence," March 2026.
* * *
The Center for AI Standards and Innovation
The Center for AI Standards and Innovation (CAISI) is a small office within the Department of Commerce that evaluates the capabilities of AI models, lends AI expertise across federal agencies, and analyzes international AI competition, such as in its 2025 report on the performance and censorship of leading Chinese and U.S. AI models.56 Most of its staff are engineers and machine learning experts with experience at frontier AI companies, research universities, and startups. It also has memoranda of understanding and non-disclosure agreements with top AI developers, including OpenAI and xAI, and with allied foreign governments.57
But CAISI lacks adequate funding, staff, and a focused mission. In its lifetime (beginning in 2023), it has only received $30 million in total funding. Analogous AI institutes in other countries, including Canada, Singapore, and the United Kingdom, have all received larger appropriations, often with smaller mandates (see Figure 2) [View figure in the link at bottom]. CAISI has only 20-30 full-time employees, which limits its capacity to fulfill its many missions and field inbound requests from other agencies like those in the Intelligence Community. It has also received a large volume of missions from the White House's AI Action Plan and a June 2025 announcement by Commerce Secretary Howard Lutnick.58 These two documents alone gave CAISI at least 21 distinct taskings.59
* * *
56 National Institute of Standards & Technology, "CAISI Evaluation of DeepSeek AI Models Finds Shortcomings and Risks," September 30, 2025.
57 The White House, "Memorandum of Understanding between the Government of the United States of America and the Government of the United Kingdom of Great Britain and Northern Ireland regarding the Technology Prosperity Deal," Presidential Memoranda, September 18, 2025.
58 U.S. Department of Commerce, "Statement from U.S. Secretary of Commerce Howard Lutnick on Transforming the U.S. AI Safety Institute into the Pro-Innovation, Pro-Science U.S. Center for AI Standards and Innovation," Press Release, June 3, 2025; The White House, "Winning the Race: America's AI Action Plan," July 2025.
59 Cole Salvador, Jack Crovitz, and Yusuf Mahmood, "Building AI Readiness in the U.S. Government," America First Policy Institute (AFPI), March 31, 2026.
* * *
Figure 2: U.S. Center for AI Standards and Innovation (CAISI) Funding Compared to Peer Institute
Caption: Committed funding (total, not annual) for the U.S. Center for AI Standards and Innovation (CAISI) lags far behind analogous institutes in other countries and remains 11 times smaller than the UK's equivalent.60
* * *
If CAISI is to help implement President Trump's AI agenda and foresee the threats posed by the CCP's AI ambitions, Congress must act to authorize it, fund it, and streamline its missions. In our report, we outline an America First vision for CAISI in which it serves the following roles: technical strike team, bridge between industry and government, frontier analysis unit, and technical standards organization.
Congress could consider allocating $50-100 million to CAISI annually. With a clarified mission and adequate funding, CAISI could become America's first line of defense against the CCP's illicit campaign to overtake America in AI.
* * *
60 Singapore Intercom Media Development Authority, "Digital Trust Centre designated as Singapore's AI Safety Institute," May 22, 2024; Government of Canada, "Canada launches Canadian Artificial Intelligence Safety Institute," November 12, 2024; UK Department for Science, Innovation, & Technology, "AI Opportunities Action Plan: One Year On," Policy Paper, January 29, 2026; Australian Assistant Minister for Science, Technology and the Digital Economy Andrew Charlton, "National AI Plan: Empowering all Australians," Press Release, December 2, 2025; U.S. Senate Committee for Appropriations, "BILL SUMMARY: Commerce, Justice, Science, and Related Agencies Fiscal Year 2024 Appropriations Bill," March 3, 2024.
* * *
The Bureau of Emerging Threats
The Bureau of Emerging Threats (ET) is an office in the State Department with similar promise and challenges to CAISI. It is "charged with anticipating and responding" to "U.S. adversaries' weaponization of advanced technology, including artificial intelligence."61 This mandate makes ET a natural candidate for analyzing the increasingly state-backed CCP campaign to acquire American AI. The Bureau could, for example, analyze trends in Chinese AI-enhanced military software like Chinese satellite intelligence reportedly being used by Iran to target U.S. forces.62 It might also analyze claims that AI threatens to undermine nuclear deterrence.63 To achieve these goals, Congress could authorize ET and provide it with direct funding. It could be mandated to deliver national security insights on emerging technology to key policymakers through regular reports.
Recommendations for the Committee
Congress has a key role to play in addressing and anticipating the CCP's offenses in the AI race. In this section, we recommend legislative actions that would help the U.S. Government counter the CCP's illicit campaign to acquire American AI technology. As a 501(c)(3), the America First Policy Institute does not support specific legislation.
Create an anti-distillation task force. Distillation attacks have become Chinese AI companies' preferred method of illegally extracting capabilities from American developers. Stopping distillation attacks would stop China's easiest path to free-riding on U.S. AI capabilities, but private American developers say that no single company can adequately defend against them alone. To combat distillation, Congress could mandate the NSA's AI Security Center coordinate with CAISI, the Bureau of Emerging Threats, other Intelligence Community offices, and industry to develop an anti-distillation task force. This task force could create and disseminate best practices for preventing distillation attacks, share threat intelligence, and conduct other relevant activities as appropriate.
Establish whistleblower protections. Congress could pass legislation to prohibit employment discrimination against whistleblowers reporting AI security vulnerabilities. Some security researchers at frontier AI companies complain that they are discouraged from informing policymakers about major security vulnerabilities.64 In particular, widespread severance and nondisclosure agreements (NDAs) at frontier labs prevent researchers from speaking out. This means that American policymakers and the national security community may not understand the seriousness of Chinese exfiltration and subversion operations until it is too late. As Senator Grassley has explained: "Today, too many people working in AI feel they're unable to speak up when they see something wrong. Whistleblowers are one of the best ways to ensure Congress keeps pace as the AI industry rapidly develops."65 Congress could consider passing legislation to protect security whistleblowers in the AI industry from retaliation by AI companies.
* * *
61 Shannon K. Kingston, "State Department launches effort to counter cyberattacks, AI risks from Iran, others," ABC News, March 23, 2026.
62 Henry Zartz, "Chinese AI satellite intelligence helping Iran target U.S. forces with 'incredible precision', analysts say," ABC News, April 6, 2026.
63 Jason Pruet, Anna Makanju, Jonathan Reiber, and Josh Achiam, "AI and International Security: Pathways of Impact and Key Uncertainties," OpenAI, February 6, 2026.
64 "OpenAI Employees Call for Protections to Speak Out on AI Risks," Bloomberg, June 4, 2024; Pranshu Verma, Kat Zakrzewski, & Nitasha Tiku, "OpenAI illegally barred staff from airing safety risks, whistleblowers say," Stars & Stripes, July 13, 2024; "OpenAI, Google DeepMind employees sign open
letter calling for whistle-blower protections to speak out on AI risks," South China Morning Post, June 5, 2024.
* * *
Establish basic security standards for frontier AI labs. Congress could pass legislation to empower the Department of War to institute minimum security standards for large AI developers in order to prevent Beijing from stealing or subverting American AI. The 2026 National Defense Authorization Act (NDAA) made a promising first step by directing the Pentagon to "develop a framework for the implementation of cybersecurity and physical security standards and best practices relating to covered artificial intelligence and machine learning technologies."66 However, the Pentagon has not yet published this framework, and the statute imposes no deadline by which the Department must do so. Given the seriousness of the Chinese security threat, Congress could build on this initiative by requiring the Department of War to develop and publish frontier AI lab security standards within a matter of months. Another weakness of the security standards authorized by the 2026 NDAA is that they apply only to AI labs that contract with the Pentagon. Given that some American frontier AI labs no longer sell AI tools to the Pentagon67, Congress could consider empowering the Department of War to impose security standards on all American AI labs regardless of their status as military contractors. These standards could include strong incident reporting requirements for attempted security breaches by nation-state adversaries.
Establish AI lab security red-teaming exercises. Congress could consider directing the NSA to lead redteaming exercises to uncover security vulnerabilities in the American AI development supply chain. NSA has deep expertise in offensive and defensive cyber operations, including classified threat vectors, so it is uniquely positioned to simulate the tactics of nation-state adversaries. Red-teaming exercises could target the full AI development process, from data scraping and filtering systems to training infrastructure to model weight storage facilities. They could also cover the full set of IP exfiltration and model poisoning attack vectors, including threats from compromised insiders. NSA's AI Security Center could collaborate with DHS and other parts of the intelligence community to ensure these exercises comprehensively simulate real-world attacks from nation-state adversaries. The exercises could be conducted in close partnership with frontier AI labs and on a fully voluntary basis.
* * *
65 U.S. Senate Committee on the Judiciary, "Grassley Introduces AI Whistleblower Protection Act," Press Release, May 15, 2025.
66 National Defense Authorization Act for Fiscal Year 2026, S. 1071, 119th Cong. (2025).
67 Cade Metz, Julian E. Barnes, and Sheera Frenkel, "Pentagon Officially Notifies Anthropic It Is a 'Supply Chain Risk'," New York Times, March 5, 2026.
* * *
Request a National Intelligence Estimate on Chinese AI. Several features of China's illicit campaign remain opaque: the extent of state support in some activities is unknown, the capability gains from different types of theft are difficult to quantify, and the intentions of Chinese leadership are unclear. In the 2026 National Defense Authorization Act, Congress required the Director of National Intelligence (DNI) to "produce a National Intelligence Estimate with respect to advancements by the [PRC] in biotechnology."68 Congress could ask DNI to produce a National Intelligence Estimate on advancements by the PRC in frontier AI. It would evaluate the country's drivers of AI progress, computing access and diversion, illicit campaign to acquire American AI, and other factors as appropriate.
Authorize and fund the Center for AI Standards and Innovation (CAISI). CAISI is a small, talent-dense technical office within the Department of Commerce that the Trump Administration and congressional leaders have recognized as a key asset in AI competition. But it lacks adequate staffing, funding, and authorization. Congress could allocate $50-100 million to CAISI annually and mandate it to act as a technical strike team, a bridge between industry and government, a frontier analysis unit, and a technical standards organization.
Authorize the Bureau of Emerging Threats (ET). ET is a new office in the Department of State with expertise in AI, military uplift, and international relations. It could complement CAISI by analyzing the strategic AI competition between the U.S. and the CCP from an international perspective and with an eye toward the future. For the Bureau to succeed, Congress could directly authorize it to deliver national security insights on emerging technology to key policymakers, such as through regular reports.
Conclusion
The United States and the Chinese Communist Party (CCP) are in a race to develop and deploy artificial intelligence. Overwhelming evidence shows that the CCP and its companies are using a suite of escalating actions to steal, subvert, and sabotage American AI. Unless we face this challenge with urgency and resolve, the CCP may succeed in its goal to achieve global AI leadership by 2030 -- an unacceptable outcome for American prosperity and power.
* * *
Biography
Yusuf Mahmood is Director of AI and Emerging Technology Policy at the America First Policy Institute (AFPI). He holds a J.D. from Harvard Law School and dual bachelor's degrees in economics and philosophy from the University of Maryland, College Park.
* * *
68 National Defense Authorization Act for Fiscal Year 2026, S. 1071, 119th Cong. (2025).
* * *
Original text and figures here: https://docs.house.gov/meetings/ZS/ZS00/20260416/119165/HHRG-119-ZS00-Wstate-MahmoodY-20260416.pdf