Artificial Intelligence and Criminal Exploitation: A New Era of Risk

Immigration Enforcement and Sanctuary PoliciesHouse Judiciary Subcommittee on Crime and Federal Government Surveillance · 2025-07-16 · 119th Congress
The House Judiciary Subcommittee on Crime and Federal Government Surveillance held this hearing to examine how criminals are exploiting artificial intelligence — through deepfakes, synthetic child sexual abuse material (CSAM), voice-cloning scams, and identity fraud — and how law enforcement can responsibly deploy AI to counter these threats. Begins at 0:23:52
Transcript
Highlights

Title

AI-driven crime, deepfakes, and law enforcement's use of AI tools

Purpose

The House Judiciary Subcommittee on Crime and Federal Government Surveillance held this hearing to examine how criminals are exploiting artificial intelligence — through deepfakes, synthetic child sexual abuse material (CSAM), voice-cloning scams, and identity fraud — and how law enforcement can responsibly deploy AI to counter these threats. Four witnesses testified on the legal gaps in current criminal statutes, the risks of AI-enabled facial recognition, and the debate over a proposed federal moratorium on state AI regulation. Begins at0:23:52

Who spoke

Chairman Andy Biggs (R-AZ)0:24:23: Opened by describing AI-enabled threats including CEO voice-cloning fraud0:25:21 and a Colorado mother scammed out of $1,000 by an AI-cloned "daughter's" voice0:26:19; later pressed witnesses on whether existing law covers malicious deepfakes of public figures made "with malice"1:15:101:16:38, and closed by asking how soon algorithmic decision-making could replace human judgment in warrants and probable cause1:38:261:39:59.

Ranking Member Lucy McBath (D-GA)0:30:06: Detailed the wrongful arrest of pregnant Detroit resident Porcha Woodruff based on a facial-recognition misidentification0:31:370:32:34; cited Detroit PD data showing 95.6% of a year's facial-recognition searches targeted Black individuals1:05:01; opposed the proposed federal moratorium on state AI regulation, noting 17 Republican governors and 40 state attorneys general opposed it and the Senate rejected it 99-10:33:570:35:26; entered both letters into the record0:36:26.

Dr. Andrew Bowne, George Washington University0:39:39: Testified AI's dual-use nature means legitimate tools can enable fraud, deepfake CSAM, and sextortion0:40:330:42:01; identified computer vision, GANs, and LLMs as the three areas of greatest concern0:41:31; flagged gaps in criminal statutes for autonomous and cross-border AI crime0:44:03; cited HR 1283 as pending legislation to extend Section 2252A (CSAM statute) to AI-generated content1:27:40.

Ms. Zara Perumal, Overwatch Data0:44:51: Said AI lowers barriers to cybercrime, helps criminals defeat identity verification, and enables highly personalized scams, including "nudifying" apps targeting children0:46:520:47:19; urged investment in education and public-private partnerships to shift technical advantage to defenders0:47:491:11:37.

Mr. Cody Venzke, ACLU0:49:08: Warned that AI regulation must not erode First Amendment and Fourth Amendment protections0:49:270:50:27; argued the proposed AI moratorium would have preempted state deepfake and privacy laws, including Tennessee's Elvis Act0:51:270:51:51; called for a moratorium on law enforcement use of facial recognition given its disproportionate impact on Black individuals1:06:11; warned federal data consolidation risks creating an unaccountable surveillance platform0:52:400:53:01.

Mr. Ari Redbord, TRM Labs0:53:56: Said criminals are consistently early adopters of new technology, from cars to crypto to AI0:54:19; cited a 456% year-over-year rise in AI-enabled scams tracked via TRM's Chainabuse platform0:56:36; said blockchain's traceability lets investigators "pinpoint" real-world actors behind wallets1:01:52; described a $225 million civil forfeiture case against a pig-butchering scam network filed roughly two weeks earlier1:37:20.

Rep. Ted Budd (R-NC)0:58:47: A former prosecutor, asked how AI and blockchain affect domestic-versus-international criminal enforcement1:00:02 and whether aggressive prosecution risks stifling domestic AI innovation1:03:07; in the second round, asked each witness how close society is to fully autonomous criminal AI activity, drawing responses that all four believe rudimentary autonomous crime already exists1:20:161:20:47.

Rep. Laurel Lee (R-FL) [0:23:52, questioning at 1:09:42]: Noted her work on the House's Bipartisan AI Task Force [0:09:42/0:10:13 in prior numbering—see below]; highlighted Operation Renewed Hope, where AI helped identify and rescue child-exploitation victims1:10:39; asked Perumal and Redbord for legislative and funding priorities to keep law enforcement ahead of AI-enabled crime1:11:071:12:31.

Rep. Robert Garcia — corrected to Rep. from California, Mr. Kelly (D-CA)1:32:42: Asked Redbord and Perumal about coordinating public- and private-sector resources for law enforcement, citing Google's recent AI-driven detection of a software vulnerability "in the wild"1:33:57; highlighted a $225 million pig-butchering forfeiture case as an example of network-level disruption1:37:20.

Key moments

Chairman Biggs cited a 456% rise in generative-AI-enabled scam reports between May 2024 and May 2025, a figure Redbord separately confirmed from TRM's own Chainabuse data0:27:180:56:36.

McBath detailed the Porcha Woodruff case: Detroit police used facial recognition to misidentify a nursing student, who was 8 months pregnant, handcuffed her, and held her 11 hours before charges were dropped a month later0:31:370:32:34.

McBath's staff data: Detroit PD's 2020 facial-recognition searches were "all" on African American individuals, rising to 95.6% the following year1:04:331:05:01.

Venzke stated the ACLU's position that law enforcement should not use facial recognition at all, citing accuracy problems with lighting, image quality, and protected classes1:06:54.

Sharp exchange between Chairman Biggs and Venzke over whether malicious deepfakes of public figures are protected speech: Venzke said such content is largely First Amendment-protected commentary on politicians, while Biggs pressed on defamation liability "with malice"1:16:071:17:03.

All four witnesses agreed, in response to Rep. Budd, that rudimentary autonomous AI-driven criminal activity already exists today, with Bowne saying "we're there"1:20:471:21:09.

Bowne noted HR 1283 would amend 18 U.S.C. §2252A to cover AI-generated CSAM but cautioned any such amendment must still satisfy First Amendment constraints1:27:401:28:23.

Redbord disclosed a $225 million civil forfeiture complaint filed roughly two weeks earlier by the US Attorney's Office in DC targeting a pig-butchering scam network1:37:20.

Venzke said the Senate rejected the proposed AI-regulation moratorium 99-1 after opposition from 17 Republican governors and 40 state attorneys general, but the House Energy and Commerce Committee chair has vowed to continue pursuing it0:33:570:35:261:30:11.

Biggs closed by asking how long before algorithms, rather than humans, are the "first mover" in decisions like probable-cause determinations; Redbord said he remains "happy with judges still making decisions around probable cause" while Bowne called it a policy choice requiring human-in-the-loop safeguards1:38:261:43:12.

Metadata

CommitteeHouse Judiciary Subcommittee on Crime and Federal Government Surveillance
Chamber / CongressHouse · 119th Congress
Date2025-07-16
TypeHearing
Witnesses
Ms. Zara Perumal — Chief Technology Officer, Overwatch Data
Mr. Ari Redbord — Global Head of Policy, TRM Labs
Dr. Andrew Bowne — Professor, George Washington University
Mr. Cody Venzke — Senior Policy Counsel, National Political Advocacy Division
Videoyoutube
Transcript182 caption blocks · 12,819 words · 1:44:18 runtime
EventCongress.gov 118467