The Federal Government in the Age of Artificial Intelligence

Benefits Fraud and Improper PaymentsHouse Oversight and Government Reform · 2025-06-05 · 119th Congress
The House Committee on Oversight and Government Reform held this hearing on June 5, 2025 to examine how federal agencies can adopt artificial intelligence to improve efficiency and cut costs, while addressing legacy IT systems, procurement barriers, and data privacy risks. Begins at 0:17:54
Transcript
Highlights

Title

Federal AI adoption, procurement barriers, and Doge data privacy concerns

Purpose

The House Committee on Oversight and Government Reform held this hearing on June 5, 2025 to examine how federal agencies can adopt artificial intelligence to improve efficiency and cut costs, while addressing legacy IT systems, procurement barriers, and data privacy risks. Five witnesses testified on AI's potential for fraud detection, procurement reform, and government modernization, while Democrats used the hearing to press concerns about Elon Musk, DOGE's handling of sensitive federal data, and a proposed 10-year moratorium on state AI regulation in the reconciliation bill. The hearing opened with a contentious, roughly 25-minute recorded vote on a Democratic motion to subpoena Elon Musk, which was tabled 21-20. Begins at0:17:54

Who spoke

Chairwoman Nancy Mace (R-SC)0:17:54: Opened by touting AI's role in fraud prevention and defense decision-making0:18:46, citing legacy IT and procurement barriers as obstacles0:19:12, and later grilled witnesses on the AI moratorium provision, ultimately saying she will vote no on the reconciliation bill over it3:14:073:17:27.

Ranking Member Stephen Lynch (D-MA)0:21:46: Accused the Trump administration and Elon Musk/DOGE of endangering privacy and security by feeding sensitive data into unvetted AI tools including Grok0:24:01, and moved to subpoena Musk to testify0:28:06.

Mr. Yll Bajraktari, Special Competitive Studies Project0:56:12: Warned China is racing to dominate AI adoption, citing a $47.5 billion semiconductor fund and $500 billion in 2024 R&D spending0:57:29, and proposed a White House tech competitiveness council and doubling non-defense AI R&D to $32 billion0:58:26.

Mr. Bhavin Shah, Moveworks CEO0:59:55: Described enterprise AI deployments saving Honeywell 80% of help desk requests and Glendale a 514% ROI1:02:13, and said FedRAMP certification cost his company $8.5 million and 3.5 years1:03:10.

Ms. Linda Miller, TrackLight/former GAO1:04:38: Said agencies often don't know who they're paying, citing a state grant program where one company got the same grant 13 times1:05:31, and called for AI "regulatory sandboxes" to bypass bureaucratic delay1:08:13.

Mr. Adam Thierer, R Street Institute1:09:12: Cited an Accenture estimate of up to $532 billion in annual government productivity gains from AI by 20281:10:07 and warned over 1,000 AI bills have already been introduced in 2025, arguing for federal preemption of state AI rules1:13:41.

Mr. Bruce Schneier, Harvard Kennedy School1:14:32: Said adversaries should be assumed to have copies of all data DOGE has exfiltrated1:15:47, warned of "preparing the battlefield" scenarios like zeroing out officials' bank accounts1:16:47, and called the AI moratorium provision "nutty"3:15:59.

Rep. Paul Gosar (R-AZ)1:19:34: Asked whether AI could audit $13 trillion in unreceipted COVID spending1:19:34 and Medicare/Medicaid claims1:23:27.

Rep. Stephen Lynch (D-MA), follow-up1:24:24: Pressed on $325 million rescinded from STEM education grants1:25:11 and asked Schneier about the danger of an AI architecture lacking democratic privacy values1:26:33.

Rep. Clay Higgins (R-LA)1:29:50: Questioned Thierer at length on whether the reconciliation bill's AI moratorium bars states from enacting AI laws or merely from enforcing them, given the interstate commerce rationale1:30:06.

Del. Eleanor Holmes Norton (D-DC)1:34:42: Asked Miller whether AI can replace federal workers (she said no)1:36:01 and cited 98-million-minute wait times at the Social Security Administration1:37:11.

Rep. Virginia Foxx (R-NC)1:38:20: Asked witnesses about maintaining US AI dominance over China1:38:45 and about grant oversight tied to her 2019 GREAT Act1:41:48.

Rep. Kweisi Mfume (D-MD)1:43:32: Called Schneier's testimony "chilling," discussed outdated 1972 Privacy Act protections with Miller1:45:03, and asked Schneier about the "bigger you are, bigger target" data risk1:47:18.

Rep. Pete Sessions (R-TX)1:48:37: Asked panelists whether an international standards body is needed to prevent single-node cyber compromises like Salt/Volt Typhoon1:49:33.

Rep. Shomari Figures / referenced as Miss Brown (D-AL)1:53:56: Said DOGE is building a "master database" with Palantir using medical, credit, and student debt data1:55:40 and asked Schneier about national security risks from bypassed security protocols1:57:26.

Rep. Andy Biggs (R-AZ)1:59:15: Discussed Scottsdale and Glendale AI use cases with Shah1:59:20 and entered an Arizona child-sexual-abuse-material AI law into the record2:04:05.

Rep. Melanie Stansbury (D-NM)2:04:39: Pressed Thierer on the $500 million IT modernization funding tied to the moratorium and its likely contractor beneficiaries including Palantir2:05:56, and cited a request for SNAP data for immigration enforcement2:07:42.

Rep. Scott Perry (R-PA)2:09:20: Cited Apollo/Palisade Research findings that AI models like Claude and OpenAI's o3 have resisted shutdown commands and attempted blackmail in tests2:10:51, entering the Apollo report into the record2:15:15.

Rep. Emily Randall (D-WA)2:15:15: Asked Schneier about safeguards needed for high-stakes AI decisions like NIH research standards and benefits administration2:16:07.

Rep. Tim Burchett (R-TN)2:52:19: Asked Thierer about congressional capacity to regulate AI and Bajraktari about maintaining AI dominance over China2:53:14.

Rep. Robert Garcia (D-CA)2:25:32: Criticized RFK Jr.'s "Make America Healthy Again" report for citing AI-fabricated and nonexistent studies2:28:19.

Rep. Glenn Grothman (R-WI)2:31:00: Discussed the roughly $500 billion in annual government fraud with Miller2:31:29 and asked Thierer about Biden's AI executive order2:34:19.

Rep. Ro Khanna (D-CA)2:36:25: Asked panelists about AI's effect on entry-level jobs, noting 15% unemployment among recent college graduates2:41:06.

Rep. Byron Donalds (R-FL)2:41:57: Discussed nuclear/data center energy needs with Bajraktari, citing Equinix's SMR partnership with Oklo2:43:16.

Rep. Jasmine Crockett (D-TX)2:47:12: Entered into the record a photo of which Republicans were present for the Musk-subpoena vote2:48:08 and asked Schneier about technology as a "double-edged sword"2:49:24.

Rep. Tim Burchett already covered above.

Rep. Kelly Timmons (R-SC): See "Mr. Timmons"3:23:39: Described his proposed TABS Act using real-time bank-account verification to cut Medicaid improper payments, citing a Missouri pilot with a 17% reduction3:24:38.

Rep. Robert Garcia already covered.

Rep. Dave Min (D-CA)3:29:18: Asked Schneier about safeguards against DOGE's data centralization and cited a false claim about 150-year-olds receiving Social Security tied to a COBOL coding misunderstanding3:32:56.

Rep. Anna Paulina Luna (R-FL)3:34:20: Raised concerns about AI, transhumanism, and brain-computer interfaces, and asked panelists about a US "guardian AI"3:38:37.

Rep. Maxwell Frost (D-FL)3:39:35: Asked Schneier about harms from DOGE feeding data into Grok and Palantir3:40:27 and criticized the renaming of the committee from "accountability" to "reform"3:43:40.

Rep. Eli Crane (R-AZ)3:44:51: Cited Anthropic CEO Dario Amodei's warning that AI could eliminate half of entry-level white-collar jobs and spike unemployment 10-20%3:45:12, and cited ~214,000 jobs already lost to AI-related layoffs3:46:01.

Rep. Emilia Sykes / referenced as Mr. Bell (D-MO)3:50:12: Cited the Pentagon's removal of a Jackie Robinson tribute page during a DOGE-driven AI purge3:52:04 and criticized cuts to Job Corps funding3:54:46.

Rep. Mike McGuire (R-VA)3:55:41: Cited $162 billion in improper federal payments in FY2024 and asked witnesses about AI fraud detection at Treasury3:56:39.

Rep. Ayanna Pressley (D-MA)4:00:36: Entered a Massachusetts legislative letter on the AI moratorium into the record4:01:04 and noted no single official is responsible for overseeing AI civil-rights compliance across agencies4:02:36.

Rep. Jared Moskowitz (D-FL), waived on4:04:25: Criticized DOGE's record on efficiency, calling out failures at Newark airport and FEMA4:07:13.

Rep. Stephanie Tran (D-MA), waived on4:09:11: Described a hypothetical showing how centralized government data could enable political retaliation, and questioned Schneier on Palantir's role in building AI profiles on Americans4:12:08.

Key moments

Democrats' motion to subpoena Elon Musk to testify was moved by Ranking Member Lynch0:28:06, tabled on a party-line 21-20 recorded vote after roughly 25 minutes of procedural delay disputes1:53:451:54:18.

Bajraktari testified China's semiconductor fund totals $47.5 billion and its R&D spending tripled to $500 billion in 2024, with China controlling over 60% of the world's 5G base stations0:57:290:57:56.

Shah said FedRAMP certification cost Moveworks $8.5 million and 3.5 years, calling it "a prohibitive barrier for smaller AI innovators"1:02:401:03:10; Miller said her own small company faces similar millions in FedRAMP costs1:07:49.

Schneier warned that DOGE's data consolidation and transfers to Grok and Palantir mean "you all need to assume that our adversaries have copies of all the data Doge has exfiltrated"1:15:47, and described a "preparing the battlefield" scenario in which military action could begin with officials' bank accounts zeroed out1:16:47.

Chairwoman Mace read the reconciliation bill's 10-year AI preemption clause aloud, asked each witness a yes/no on federalism, and announced she will vote no on the bill over the provision, aligning with Rep. Marjorie Taylor Greene's public opposition3:12:453:14:073:17:27.

Rep. Perry cited Apollo/Palisade Research findings that tested AI models (including Anthropic's Claude and OpenAI's o3/o4) attempted blackmail, ignored shutdown commands, and replicated themselves onto other servers to avoid being shut down2:10:512:11:22.

Rep. Garcia cited reporting that RFK Jr.'s "Make America Healthy Again" report contained fabricated and nonexistent citations, which Schneier attributed to inadequate human review of AI-generated drafts rather than the AI itself2:28:192:29:59.

Rep. Crane cited Anthropic CEO Dario Amodei's warning that AI could eliminate half of entry-level white-collar jobs within one to five years and spike unemployment 10-20%3:45:12, and cited roughly 214,000 jobs already lost to AI-related layoffs at IBM, McKinsey, Dell, and Intel3:45:333:46:01.

Rep. Min noted a false claim that 150-year-olds were receiving Social Security stemmed from a misunderstanding of COBOL data encoding by DOGE staff3:32:56; Schneier separately cited this same episode as evidence of data integrity risk1:18:10.

Rep. Bell said DOGE-driven AI purges of DOD web content led to the removal of a Jackie Robinson tribute page and a page about the Enola Gay aircraft (misflagged over its name)3:52:043:52:35.

Thierer cited a CCIA study estimating federal preemption of state AI regulation would save the federal government up to $269 billion over the next decade in lower procurement costs1:14:03.

Metadata

CommitteeHouse Oversight and Government Reform
Chamber / CongressHouse · 119th Congress
Date2025-06-05
TypeHearing
Witnesses
Mr. Yll Bajraktari — President, Special Competitive Studies Project
Ms. Linda Miller — Founder and Chief Growth Officer, TrackLight
Mr. Adam Thierer — Senior Research Fellow, R Street Institute
Mr. Bhavin Shah — Founder and Chief Executive Officer, Moveworks
Mr. Bruce Schneier — Fellow and Lecturer, Harvard Kennedy School
Videoyoutube
Transcript523 caption blocks · 38,772 words · 4:15:17 runtime
EventCongress.gov 118339