The Capacity Gap: Why Crime Fighters Are Falling Behind | #cybercrime | #infosec

Introduction

If you think of the word “crime,” what comes to mind? For most of us, it’s burglars, pickpockets, and car thieves — criminals operating in the physical world. The biggest and fastest-growing threats, however, are digital.

In 2000, the FBI launched the Internet Crime Complaint Center (IC3) to track cybercrimes. Since 2020, the monetary losses reported to the IC3 have increased 5x to a total of USD 20.9 billion (2025), with a 27% increase in total complaints, while in that same time period, reported property crimes nationwide have fallen nearly 20%.¹

Crime fighters have responded aggressively to this threat. Law enforcement at all levels — federal, state, and local — has frozen and seized more scam proceeds than ever: the FBI’s Financial Fraud Kill Chain, which coordinates with banks to freeze funds lost to fraud, froze a record USD 679 million in 2025, up 21% from the previous year. The Justice Department’s Scam Center Strike Force, created in 2025, has already frozen or seized nearly USD 1 billion in cryptocurrency stolen through scams. And while many state and local agencies lack cybercrime personnel or dedicated units, they have taken on the problem as well: in 2025, Virginia’s Loudoun County Sheriff’s Office seized USD 1.4 million taken in a cryptocurrency investment scam.²

Yet overall, the problem is growing faster than law enforcement can respond. From 2023 to 2025, victims reported USD 50 billion in losses to digital crime. We estimate that federal law enforcement agencies froze or recovered 4% to 10% of that total — and given that most victims do not report their losses to federal authorities, that range likely understates the problem. While IC3 reports count USD 50 billion in losses over three years, a recent Gallup survey estimates that US adults lost at least USD 68 billion to scams in 2025 alone.³

Estimated share of digital crime losses frozen or recovered by federal law enforcement: 4% to 10%

While these figures are troubling, they do not reflect a failure of our law enforcement. The problem, instead, is structural in nature. Frontier technologies, cryptocurrency among them, enable criminals to move money out of the reach of the law, before agencies can recover it — and now AI is empowering fraudsters to move even faster, and defraud victims of even more money. Today’s law enforcement does not have access to the same tools or resources; victim reports are piling up, while agency staffing and budget stay flat.

IC3 complaints vs. federal investigative positions and budget

The short-staffing, budgeting, and imbalances in tooling are all symptoms of a bigger, systemic problem: the capacity gap.

The capacity gap

The capacity gap is the difference between the amount of crime a society experiences, and the amount law enforcement officials can investigate or prevent. The gap has been growing for years due to criminals’ rapid adoption of new technology, and AI is accelerating it further.

{{premium-content_chapter-divider}}

The capacity gap is a structural disadvantage for law enforcement, and AI is widening it further

The capacity gap gives digital criminals fundamental advantages over law enforcement across four key dimensions: speed, scale, cost, and the level of expertise crime requires.

A foundational problem is that criminals and law enforcement operate under different constraints. Criminals adopt new technology the moment it’s available, work from any country, and answer to no one. In their pursuit of money, they can act with abandon and no concern for collateral damage. Law enforcement, on the other hand, must work within a defined jurisdiction, acquire tools through public procurement, and ensure that every investigation holds up in court. Of course, some of these are necessary boundaries; others are outside of law enforcement’s control, and would require legislative or regulatory involvement to change.

As it stands now, the status quo enables criminals to embrace new tools faster than law enforcement can.

AI makes that disadvantage even more dangerous, because while earlier technological advances augmented crime, AI automates it. David Agranovich, who led threat disruption at Meta for eight years, described the result in September 2026:

{{36-capacity-gap-report-quote-1}}

AI is drastically increasing criminals’ speed and scale, while eliminating the barriers of cost and expertise that previously limited their actions. Law enforcement has not kept up on these four dimensions, as the examples below quantify.

The capacity gap threatens public safety, and these figures show that the differences are too large to be closed by hiring surges or budget increases. The only way to close a gap this size is to multiply what each investigator can do with the best tooling available.

{{premium-content_chapter-divider}}

How we close the gap: AI for law enforcement and private sector collaboration

The first step to closing the capacity gap is to empower law enforcement to use AI to increase the effectiveness and scope of their work, just as criminals have. If deployed correctly, AI can act as a force multiplier at every step of the investigative pipeline for digital fraud, by:

• Triaging every scam complaint as it arrives;
• Automatically tracing and flagging funds that can still be frozen;
• Instantly assembling the open-source intelligence and financial data an investigator might spend days collecting;
• Linking hundreds of complaints from different jurisdictions at the federal, state, and local levels into broader criminal networks; and
• Enabling detectives without specialist training to benefit from the tradecraft of veteran investigators, in order to successfully trace funds and take down criminal networks.

Today’s law enforcement has the skill to successfully fight crime. AI enables them to do it faster and at broader scale.

The second step is to facilitate better partnership between the public and private sectors. Both sides are essential: only law enforcement can compel evidence, seize assets, and arrest and prosecute. The private sector brings intelligence at commercial scale, from financial data to corporate registries to the dark web; build cycles measured in weeks rather than years; and the ability to disrupt on private infrastructure, whether that means banning fraudsters’ social media accounts, shuttering malicious web domains, or automatically warning customers about a potential scam via their banking and fintech apps.

The FBI’s Operation Level Up shows the power of this kind of collaboration. Launched in January 2024 with support from the Secret Service, Level Up uses data from IC3 complaints and private sector partners to identify and alert victims of cryptocurrency investment fraud. Since launch, Operation Level Up has notified more than 10,000 victims — most of whom did not yet know they were being scammed — and saved an estimated USD 660 million, while flagging thousands of fraudster accounts to the private sector partners who could take them down.¹⁰ By applying this model across more agencies — federal, state, and local — and using AI to identify more actionable data even sooner, law enforcement can build on Operation Level Up’s success to help more victims and facilitate more criminal takedowns.

The capacity gap is not a failure of crime fighters. It is the natural result of fighting AI-powered crime with tools built for an earlier era. Agency leaders and policymakers can tip the scales by investing in AI that expands the capacity of today’s investigators, and by building the public-private partnerships that will make these expansions as effective as possible.

Law enforcement has the skill. What they need now is more capacity.

{{premium-content_chapter-divider}}

About TRM Labs

TRM Labs is the operating system for modern crimefighters, built to detect and disrupt the criminal networks that exploit frontier technologies, including cryptocurrency and AI. Combining proprietary intelligence, AI-native investigations software, and disruption network infrastructure, the platform is built for high-consequence environments where accuracy, auditability, and security are essential. TRM is trusted by 600+ government agencies and financial institutions across 75 countries to counter fraud, scams, cyber crime, child exploitation, money laundering, and sanctions evasion, among other illicit activities.

To learn more, visit www.trmlabs.com.

{{premium-content_chapter-divider}}

Sources and methodology

Most figures in this report are taken directly from the cited government, court, academic, and industry sources. Some figures are estimates using inputs from one or more source documents—in those cases, the relevant endnote for the figure provides the sourced inputs and estimation methodology.

1. Digital crime: complaints received by the FBI’s Internet Crime Complaint Center, 791,790 in 2020 (2020 Internet Crime Report) and 1,008,597 in 2025 (2025 Internet Crime Report), up 27%. Reported losses rose from USD 4.2 billion to USD 20.9 billion. Property crime: the FBI’s estimated number of property crimes nationwide (burglary, larceny-theft, and motor vehicle theft), 6,452,038 in 2020 (FBI, 2020 crime statistics) and 5,245,768 in 2025 (FBI, Summary of Reported Crimes in the Nation, 2025, p. 8), down 18.7%.

2. Losses are those reported to the FBI’s Internet Crime Complaint Center (IC3) in calendar years 2023 to 2025: USD 12.5 billion, USD 16.6 billion, and USD 20.9 billion, a total of USD 50.0 billion (2023, 2024, and 2025 Internet Crime Reports). IC3 records reported losses, but those figures don’t account for underreporting.

The lower bound estimate for funds frozen, 3.6%, counts only those frozen through the FBI’s Financial Fraud Kill Chain, which is initiated directly on IC3 complaints: USD 538.4 million in 2023, USD 561.6 million in 2024, and USD 679.0 million in 2025, USD 1.78 billion in total (Recovery Asset Team sections of the IC3 reports).

The central estimate, 7.9%, adds the Secret Service’s reported recoveries for financial crime, which include asset forfeiture, returned payments, and other losses recovered in its criminal investigations: USD 1.11 billion in fiscal 2023, USD 330 million in fiscal 2024, and USD 750 million in fiscal 2025, USD 2.19 billion in total (Department of Homeland Security Annual Performance Report, fiscal 2025 to 2027). These recoveries come from the same kinds of cases IC3 records cover but are not tied to specific complaints. These recoveries may cover law enforcement action on thefts that occurred before the start of the analysis, which would artificially inflate the central estimate’s rate of freezing and recovery.

The upper bound, 10.3%, adds federal seizures in cases that may overlap with the digital crime loss reporting from the covered time period, but whose crime category and theft year cannot be confirmed from the published data. This includes the USD 401.7 million in cryptocurrency that the Scam Center Strike Force reported as already seized and forfeited when it was announced in November 2025, some of which may also be counted in the Secret Service figure (US Attorney’s Office for the District of Columbia); Homeland Security Investigations’ virtual currency seizures of USD 148 million in fiscal 2023 and USD 192 million in fiscal 2024 (ICE annual reports: FY2023, archived copy, and FY2024; no fiscal 2025 report has been published); and USD 428.3 million seized from 2023 to 2025 in fraud, computer crime, identity theft, and money laundering cases with anticipated victims, recorded by the FBI, US Attorneys, and other agencies in the Justice Department’s Consolidated Asset Tracking System (“Consolidated Asset Tracking System (CATS) ZIP File”). The Strike Force’s total has since grown to USD 938.5 million frozen or seized as of September 15, 2026 (Strike Force program page), but only the amount reported at launch falls within the 2023 to 2025 window. An extremely generous count that also includes IRS Criminal Investigation’s cybercrime seizures (USD 925.7 million in fiscal 2024 and USD 149 million in fiscal 2025, see IRS-CI annual reports for 2024, p. 13; and 2025, pp. 8-9), whose identified cases are mostly related to darknet markets, tax crime, and pre-2020 thefts of cryptocurrency, and USD 238.8 million in FBI seizures in conspiracy and securities cases recorded in the Consolidated Asset Tracking System would push the estimate to 12.9%.

State and local agencies: the Loudoun County Sheriff’s Office in Virginia seized approximately USD 1.4 million in cryptocurrency stolen in a cryptocurrency investment (“pig butchering”) scam, which it described as one of the first and the largest seizures of its kind in Virginia (NBC Washington, February 24, 2025). Cybercrime personnel: in 2020, 23.2% of local police departments serving fewer than 100,000 residents had any personnel designated to address cybercrimes and 3.8% had a full-time specialized unit, while 78.9% of departments serving 100,000 or more residents had any personnel designated for cybercrime and 43.3% had dedicated units (Bureau of Justice Statistics, Local Police Departments: Personnel, 2020, Table 13); 72.9% of primary state law enforcement agencies (state police and highway patrols) had designated cybercrime personnel and 60.4% had a full-time specialized unit (Bureau of Justice Statistics, Primary State Law Enforcement Agencies: Personnel, 2020, Table 10); and 26.1% of sheriffs’ offices employing fewer than 100 full-time-equivalent sworn officers had any personnel designated to address cybercrime and 2.6% had a full-time specialized unit, while 82.8% of sheriffs’ offices employing over 100 officers had any designated personnel, and 38.7% had a full-time specialized unit (Bureau of Justice Statistics, Sheriffs’ Offices Personnel, 2020, Table 12).

3. Gallup and the Stop Scams Alliance, United States of Scams (2026), based on a survey of 5,173 US adults. Reporting: only 13% of 2025 scams were reported to the FTC or federal law enforcement (p. 20). Losses: an estimated USD 68 billion lost in 2025 by US adults who were personally scammed (p. 9), calculated as the share of adults personally scammed (5.8%) multiplied by the adult population (267.2 million) and the mean reported loss (USD 4,371) (p. 35).

4. Complaints: the number of complaints the FBI’s Internet Crime Complaint Center received each calendar year, from its annual Internet Crime Reports (2019: 467,361; 2020: 791,790; 2021: 847,376; 2022: 800,944; 2023: 880,418; 2024: 859,532; 2025: 1,008,597). Staff positions: FBI total direct authorized positions, enacted, all missions, compiled from the Department of Justice Budget and Performance Summaries, FBI chapters: FY2021 (FY2019: 35,416; FY2020: 35,534), FY2023 (FY2021: 35,742), and FY2026 (FY2023: 37,000; FY2024: 35,967; FY2025: 35,092); the FY2022 enacted figure is not publicly available and so was interpolated. Secret Service total positions and positions in the Homeland Security Investigations (HSI) program of ICE Operations and Support, enacted, from the Department of Homeland Security Congressional Budget Justifications: Secret Service FY2021 (FY2019: 7,650; FY2020: 7,777), FY2022 (FY2021: 7,896), FY2024 (FY2022: 8,105; FY2023: 8,305), and FY2026 (FY2024 and FY2025: 8,382); HSI FY2021 (FY2019: 8,778; FY2020: 8,784), FY2022 (FY2021: 8,856), FY2024 (FY2022: 8,965; FY2023: 8,602), and FY2026 (FY2024 and FY2025: 8,869). Combined positions: 51,844 in FY2019 and 52,343 in FY2025. Budget: FBI Salaries and Expenses budget authority received from the DOJ summaries and Congressional Research Service overviews of Commerce, Justice, Science appropriations (FY2020, FY2021 and FY2022, FY2023, FY2024): USD 9,192 million in FY2019 rising to USD 10,644 million in FY2025, up 16% in nominal terms. Secret Service Operations and Support and the HSI program, enacted, from the same budget justifications: Secret Service USD 2,148.5 million in FY2019 and USD 3,008.0 million in FY2025; HSI USD 1,916.2 million and USD 2,459.1 million. Combined budget for the three agencies: USD 13,257 million in FY2019 and USD 16,111 million in FY2025, up 22% in nominal terms. FY2025 figures for the Department of Homeland Security agencies are full-year continuing resolution levels. All budget figures deflated with the BLS CPI-U annual average (255.7 in 2019, 321.9 in 2025).

5. David Agranovich, post on X, September 10, 2026, responding to Anthropic’s threat intelligence report published that day. Agranovich led threat disruption at Meta from 2018, as Director of Global Threat Disruption from 2021 to 2026.

6. Scam calls figure: Voice-Enabled AI Agents can Perform Common Scams (2024), Table 1. Autonomous voice agents in a test environment completed common scams in 77 seconds (stealing a login) to 183 seconds (talking a victim through a bank transfer); 130 seconds taken as midpoint, with a 36% success rate. Charging decisions: Bureau of Justice Statistics, Federal Justice Statistics 2023, Table 4. Fraud matters concluded by US attorneys in fiscal 2023 waited a median of 576 days from referral to the decision whether to charge. Note that this figure does not include the investigation before referral and any court time, which would add more days.

7. State costs to prosecute fraud: Home Office, Economic and Social Cost of Fraud 2023 to 2024, Table 5: GBP 422 million spent responding to fraud against individuals (GBP 179 million law enforcement, GBP 243 million criminal justice), divided by the 3,699 fraud offenses charged or summonsed in the year to March 2024 (Home Office, Crime Outcomes in England and Wales, data tables for year March 2025, Table 6.2.1, year ending March 2024 column). GBP 114,085 per charge, converted USD 144,900 at 1.27 exchange rate. Scam compound costs per successful fraud: a modelled profit-and-loss for one large Southeast Asian compound, USD 5,989 per victim defrauded (payroll USD 3,229; protection and bribes 1,035; management 621; laundering 414; premises 310; recruitment and trafficking 193; utilities 124; technology 62), built from UNODC reporting on recruitment offers (Casinos, Cyber Fraud, and Trafficking in Persons for Forced Criminality in Southeast Asia, 2023, pp. 11 to 12: pay offers of USD 1,000 to 1,500 a month, USD 300 paid per person recruited, about USD 2,500 paid to a recruitment agency per person), congressional testimony on China-based professional money laundering services placing fees at 1-2% of value laundered (Christopher Urben, testimony before the House Committee on Oversight and Accountability, April 26, 2023, p. 1), and reporting on worker pay of about USD 800 a month plus commission (Viet Nam News), April 21, 2026, citing police on a Cambodia-based ring. Protection and official payments are set at 5% of revenue, modeled on fixed monthly protection fees of USD 30,000 to 50,000 that UNODC reports illicit online operators in the region pay to intermediaries (Transnational Organized Crime Threat Assessment for South-East Asia 2026, p. 134), set against monthly team targets of about USD 1 million reported from leaked compound records (Wired, January 27, 2026). Management is set at 3% of revenue, based on figures recorded in a cross-border fraud case published by China’s Supreme People’s Court on July 26, 2024. Premises are set at USD 1,500 per worker per year, based on the range modeled from reported scam compound leasing costs (Reuters, July 14, 2026).

8. Home Office, Crime Outcomes in England and Wales, year ending March 2025: 1,225,778 fraud offenses recorded by Action Fraud, Cifas, and UK Finance; 50,715 referred to a police force for investigation; 5,530 charged or summonsed (data tables, Tables 6.1 and 6.2.1). Referrals and outcomes recorded in a year can relate to offenses recorded in earlier years, because investigations can take months or longer to complete.

9. Anthropic, Disrupting the first reported AI-orchestrated cyber espionage campaign, November 2025: a Chinese state-sponsored group used AI agents to carry out 80 to 90% of an intrusion campaign against about 30 organizations, with human operators intervening at four to six decision points per campaign. Anthropic, Detecting and countering misuse of AI: September 2026: the same operating model “has now proliferated across every class of actors we investigated,” including a criminal group in which “AI agents performed nearly all of the work” and a single actor who attacked about thirty companies in about four days.

10. FBI and US Secret Service, Operation Level Up fact sheet used in reporting by Fox News (September 2026), p. 1: 10,406 victims notified since launch, 77% of whom did not know they were being scammed; USD 660 million saved; 7,670 scam accounts passed to the private sector for action. Background: Federal Bureau of Investigation, 2025 Internet Crime Report, p. 20 (“Positive Impacts”): Operation Level Up, launched in January 2024 with the support of FBI and US Secret Service agents, identifies victims of cryptocurrency investment fraud using IC3 complaint data and notifies them. Private-sector role: Patrick Wyman, chief of the FBI’s Virtual Assets Unit, TRM Talks, April 9, 2025: “by leveraging blockchain analytical tools and blockchain intelligence, what we’ve been able to do is look at transactions that are happening on the blockchain and trace those back to victims themselves,” then “working with the private sector to identify who potential victims are and then actually doing direct outreach to those victims.”