- Published on
The Compounding Pressures on American Courts
Approaching the end of my second year as Director of the Administrative Office of the U.S. Courts (AO), I can’t recall a time when the branch faced so many serious and urgent concerns. These include insufficient funding, relentless cyberattacks on our systems, threats of impeachment and physical harm against judges, and the shortage of judges after a two-decade drought in the creation of new judgeships.
A Pattern Worth Investigating
In October 2026, we published an article describing what we called the “Underinvestment Crisis in the Courtroom.” The idea grew out of conversations with court clerks, administrators, and technology teams. Different jurisdictions describe different problems, but we keep hearing variations on the same theme: difficulty finding and retaining staff, more work falling on those who remain, and new technologies that sometimes make court operations more complicated rather than less. [1]
That left us with a question. Are these separate pressures arriving at the same time, or are they interacting in self-fulfilling ways that compound their impact?
We began looking at the available data—not to prove that a single explanation accounts for every operational difficulty, but to test whether the patterns we encounter in the courtroom appear elsewhere in the judicial system. Some datasets describe federal courts while the operational experiences that prompted this work often come from state and local courts. Measures of compensation are imperfect proxies for hiring difficulty. And none of these trends, by themselves, establish cause and effect.
Those limitations matter. So, what does the evidence tell us? That courts are trying to meet evolving responsibilities with personnel, judicial capacity, and technology systems that cannot always adjust at the same pace. Our working hypothesis is that chronic underinvestment in institutional capacity makes these pressures more likely to reinforce one another.
Scope and method. This is an exploratory, research-informed industry assessment, not a representative survey or causal study. Unless identified otherwise, federal caseload data refer to federal fiscal years. State-court experience should not be inferred directly from federal trends.
Start with the people doing the work
It is easy to discuss the judiciary in terms of judges, filings, and decisions. But much of the work that makes a courtroom function happens before and after a judge enters the room. Clerks and other court personnel manage records, scheduling, filings, notices, evidence, public interactions, and the practical choreography of hearings. In some courts, these same employees also support audio recording, virtual participants, and courtroom technology. The particulars differ by jurisdiction; the dependence on capable personnel does not.
Compensation is one useful place to begin. In our initial analysis, we assembled Bureau of Labor Statistics wage series for two occupational categories: “court, municipal, and license clerks,” and “judicial law clerks.” Relative to the income benchmark used in that analysis, both series suggest a deterioration in comparative pay over roughly the past decade. The judicial law clerk series shows a particularly pronounced decline. [2]
That is not the same as demonstrating a decline in real wages, nor does a national wage comparison prove that a particular court is underpaying its employees. The first category includes municipal and license clerks who do not work in courts, and the benchmark mixes workers with different hours and employment circumstances. The appropriate conclusion is narrower—and still significant: the figures raise a serious question about whether court employment is remaining competitive in the labor market.
Figures A and B illustrate this shift over the past decade. Both occupational categories show a decline in compensation relative to the selected national income benchmark, with judicial law clerks experiencing the more pronounced change. The charts do not tell us how many qualified candidates have declined court employment, but they raise an important question: as court responsibilities continue to expand, are the financial incentives for pursuing these positions keeping pace?
Figure A. Relative Compensation of Court and Related Clerks, 2015–2025
Average annual wages for court, municipal, and license clerks, expressed as a multiple of the U.S. median individual income.
Figure B. Relative Compensation of Judicial Law Clerks, 2015–2025
Average annual wages for judicial law clerks, expressed as a multiple of the U.S. median individual income.
There is reason to take that question seriously. In its 2026 survey reporting, the National Center for State Courts describes persistent staffing shortages, particularly among clerks and clerk staff, alongside increasing case complexity and workload pressures. That is independent evidence of a problem whose effects are visible well beyond any one wage series. [3]
And a vacancy rarely means only an empty chair. Work must be reassigned, a hearing may require more preparation, and experienced employees have less time to train newcomers. Over time, the extra load can make retention harder still. It is the kind of cycle that is easy to describe and remarkably difficult to break.
The question, then, is not simply how many employees courts can hire. It is how much work the available employees are being asked to absorb—and whether the systems around them are helping.
Case backlogs are a capacity question, not a single trend line
The COVID-19 pandemic provided a severe test of judicial continuity, but it did not create every problem the courts encountered. Hearings were postponed, access to proceedings had to be reconsidered, and courts adapted at extraordinary speed. Some of the consequences lasted well beyond the emergency itself.
One of the most visible concerns was the growing number of unresolved cases. As proceedings were delayed or rescheduled, courts faced the challenge of maintaining existing workloads while accommodating new filings. But understanding what happened to those caseloads requires looking beyond a single statistic.
Federal civil caseloads are generally measured through three related figures: cases filed, cases terminated, and cases remaining pending. Each tells us something different about the judicial system's workload and its ability to process cases. Figures C, D, and E illustrate why these distinctions matter.
Figure C examines annual federal civil case filings, Figure D tracks pending civil cases, and Figure E shows annual case terminations. Together, they reveal a striking disruption beginning in 2020, followed by several years of pronounced fluctuations.
The onset of COVID-19 coincides with one of the most dramatic changes in the data. Filings rose sharply in 2020, while pending cases increased substantially and remained elevated in subsequent years. Yet terminations did not immediately rise at a comparable rate, contributing to an imbalance between the number of cases entering the system and those being resolved.
Of course, not every fluctuation in federal caseloads can be attributed directly to the pandemic. Large consolidated proceedings and changes in litigation activity can significantly influence national statistics. Nevertheless, COVID-19 introduced an extraordinary disruption to the everyday operation of courts, exposing how quickly existing processes could be overwhelmed by circumstances beyond their control.
Hearings were postponed, schedules were rearranged, and courts were forced to adapt to rapidly changing conditions while continuing to administer justice. Even when proceedings could resume, the administrative work associated with rescheduling, coordinating participants, and managing delayed cases did not simply disappear.
And while the figures show considerable improvement in pending caseloads in later years, the recovery was neither immediate nor uniform.
What makes this especially interesting is that the consequences of a disruption are not necessarily confined to the period in which it occurs. A temporary shock can create operational pressures that persist long after the original event has passed.
This also highlights an important limitation of national caseload statistics. An overall decline in pending cases does not necessarily tell us whether individual courts are becoming more efficient, whether their workloads have become easier to manage, or whether the operational pressures facing court personnel have disappeared.
And these are federal figures. State courts, which handle the overwhelming majority of litigation in the United States, operate under different structures, funding arrangements, and caseload demands. Research from the National Center for State Courts also points to the importance of staffing pressures, case complexity, and local conditions when evaluating judicial workloads. [3, 5]
Figure C. Federal Civil Cases Filed, 2015–2025
Annual civil case filings in U.S. federal courts, normalized per 1,000 U.S. residents.
Figure D. Federal Civil Cases Pending, 2015–2025
Civil cases pending in U.S. federal courts at the end of each reporting period, normalized per 1,000 U.S. residents.
Figure E. Federal Civil Cases Terminated, 2015–2025
Annual civil case terminations in U.S. federal courts, normalized per 1,000 U.S. residents.
In other words, "how many cases?" is only one part of the problem.
Consider two proceedings. One may involve a relatively straightforward dispute requiring limited evidence and a single hearing. Another may involve extensive digital evidence, multiple parties, remote participants, interpreters, and several scheduling changes. Both count as individual cases, but the administrative resources required to manage them can differ enormously.
This is where the relationship between caseloads and staffing becomes particularly important.
Even when overall case inventories decline, courts may still face increasingly complicated workflows. Scheduling, records management, evidence handling, communication with participants, and coordination among judicial personnel all require time. These responsibilities do not necessarily diminish simply because fewer cases remain pending.
There is also the question of institutional resilience.
Systems need room to recover from unexpected disruptions. When a courtroom calendar is already operating at capacity, a single postponed proceeding can trigger additional scheduling work, notifications, coordination among participants, and subsequent delays. Multiply those disruptions across an already stretched workforce, and even relatively small interruptions can have substantial consequences.
Maximizing day-to-day utilization may appear efficient, but a court with no spare operational capacity can be surprisingly fragile when something goes wrong. The pandemic demonstrated how quickly circumstances can change. While the federal data shows that courts have made meaningful progress in reducing pending civil cases, it also illustrates how dramatically workloads can fluctuate.
The larger question is therefore not simply whether courts are keeping up with today's filings. It is whether they have the personnel, infrastructure, and institutional flexibility necessary to respond when tomorrow's demands look very different.
And that question brings us to another important constraint on judicial capacity: the judges themselves.
Judgeships: another constraint on institutional flexibility
Staffing pressures do not stop at the clerk's office. Judges face many of the same institutional challenges, albeit with a different set of responsibilities. As caseloads fluctuate and proceedings grow more complex, courts must find ways to maintain the judicial capacity necessary to hear disputes, evaluate evidence, and deliver reasoned decisions.
Yet unlike many other public institutions, federal courts cannot simply create additional judicial positions when demand increases. The number of authorized Article III judgeships is determined by Congress, and significant expansion has been relatively rare. The last comprehensive expansion of federal judgeships occurred in 1990, more than three decades ago.
This raises an interesting question: If the number of authorized judgeships has remained relatively unchanged while the country has continued to grow, how have federal courts maintained their capacity?
Figures F, G, and H help illustrate the answer.
Figure F examines the number of serving Article III judges relative to the U.S. population, while Figure G shows the number of congressionally authorized judgeships using the same benchmark. Despite fluctuations in the number of serving judges, the number of authorized positions per million residents has declined considerably.
Figure H introduces another important dimension: the growing proportion of judges serving in senior status.
Under Article III, federal judges hold lifetime appointments. Eligible judges may elect to assume senior status while continuing to hear cases, often with reduced workloads. This allows courts to benefit from their experience even after they have stepped away from regular active service.
As illustrated in Figure H, the proportion of serving Article III judges holding senior status increased from approximately 30% to 40% over the period examined—a relative increase of roughly one-third.
Figure F. Active Article III Judges Relative to U.S. Population, 2001–2025
Number of serving Article III judges per million U.S. residents, including judges in senior status where counted in the underlying dataset.
Figure G. Authorized Article III Judgeships Relative to U.S. Population, 2001–2025
Congressionally authorized Article III judicial positions per million U.S. residents.
Figure H. Senior Article III Judges as a Share of Serving Judges, 2001–2025
Percentage of serving Article III judges holding senior status.
In other words, the continued participation of senior judges has helped federal courts maintain judicial capacity despite limited growth in authorized positions.
This is a valuable arrangement, but it raises questions about long-term sustainability. Senior judges make significant contributions to the judiciary, yet their participation and workloads can vary. Courts cannot assume that every judge entering senior status will continue hearing cases at the same rate, or indefinitely.
Nor is population alone a perfect measure of judicial demand. Different courts experience different caseloads, case complexity, geographic challenges, and administrative requirements. A court handling fewer but substantially more complex proceedings may face greater demands than one processing a higher volume of relatively straightforward matters.
The federal judiciary has recognized these pressures. In 2025, the Judicial Conference recommended the creation of 71 additional Article III judgeships, including 69 district court positions and two appellate positions. The broader concern, then, is not simply how many judges the country has. It is whether judicial capacity can adapt as the demands placed upon the courts change.
And this brings us back to the people supporting those judges. When judicial capacity becomes constrained, administrative pressures rarely remain confined to the bench. Scheduling, case preparation, continuances, records management, and coordination among courtroom participants can all become more difficult when court calendars are stretched.
Much like the staffing challenges discussed earlier, this is not necessarily a problem that can be solved by asking existing personnel to work harder. It is another example of why long-term institutional capacity requires sustained attention rather than temporary adjustments.
The pandemic accelerated modernization—and added another job to the courtroom
Now consider what happened to the courtroom itself. Before 2020, a given courtroom might have relied on a digital audio recording system, a case-management system, and perhaps an evidence presentation system. There were already integration and support challenges, but the basic workflows were comparatively familiar.
During the pandemic, video conferencing moved from optional capability to an operational necessity for many courts. That shift protected access and continuity. It also introduced a new layer of work: virtual waiting rooms, participant admission, remote witness management, breakout conversations, audio routing, camera controls, scheduling links, recording decisions, privacy settings, and the preservation and retrieval of records.
None of those functions is unreasonable. The difficulty appears when each function is considered in isolation. A clerk may have to operate a courtroom touch panel, manage a conferencing application, monitor a separate recording system, reconcile a case-management calendar, and then make sure the resulting files can actually be located and retained. Even reliable systems can become collectively cumbersome when the workflow connecting them is missing.
Technology has real potential to reduce workload. Yet every additional device, credential, configuration, vendor relationship, and exception procedure also creates something that someone must understand, maintain, and troubleshoot. When staffing is already tight, those additional duties are not a small detail. They are part of the investment decision. For instance, a general-purpose meeting platform can be an excellent communication tool but does not replace a complete courtroom workflow.
The answer is not to abandon modernization. It is to ask a more difficult question before purchasing it: once this system is installed, whose work becomes easier, whose work changes, and what new work appears?
Speech-to-text: a powerful tool, but no substitute for preparation
If there is one technology that seems particularly well-positioned to relieve some of the pressures facing court staff, it is speech-to-text (STT).
The appeal is obvious. Automated transcription can make recorded proceedings searchable, assist with note-taking, accelerate transcript preparation, and reduce the time staff spend reviewing hours of audio. In an environment where recruiting and retaining qualified personnel is already difficult, these capabilities represent a meaningful opportunity.
But there is a danger in treating speech-to-text as a solution to staffing shortages rather than a tool to support the people already doing the work.
For years, much of the conversation around STT centered on accuracy. Was the technology sufficiently advanced to be useful in a courtroom? Would it ever become reliable enough to justify the investment?
Those questions are becoming less relevant. Modern speech-to-text engines are already remarkably capable, and the availability of sophisticated open-source models has lowered the barrier to adoption considerably. Yet one fundamental limitation remains: speech-to-text can only work with the audio it receives.
Consider an ordinary court proceeding. An attorney turns away from the microphone while addressing the judge. A witness speaks quietly, several participants interrupt one another, or someone accidentally mutes a microphone. Background conversations, unfamiliar terminology, regional accents, and even an unfortunately timed cough can further complicate transcription.
Some of these challenges can be mitigated through better speech recognition models, microphone placement, acoustic treatment, or recording practices. Others cannot. A transcription engine cannot reliably recover words that were never captured in the first place.
This distinction matters because courts may be tempted to evaluate STT primarily by the quality of a demonstration. But the performance of a transcription engine under ideal conditions tells us relatively little about how well it will perform in an actual courtroom, particularly one with inconsistent microphone use or poorly configured audio infrastructure.
The quality of the transcription depends as much on the preparation of the environment as it does on the sophistication of the technology. And the responsibilities do not end with capturing good audio.
Courts must still determine how transcripts will be reviewed, how speakers will be identified, how the output will be associated with official recordings and case records, and how errors will be corrected. They must also consider confidentiality, retention requirements, and whether the resulting text is appropriate for its intended purpose.
A searchable draft transcript used to locate testimony is a very different proposition from an official transcript relied upon in legal proceedings. The standards of accuracy and verification must reflect that distinction.
Then there is the question of cost.
Consumption-based cloud transcription services can offer an attractive entry point, particularly when they require little upfront investment. But processing charges, storage, network dependencies, ongoing administration, and repeated transcription can introduce expenses that are difficult to anticipate without a clear understanding of usage.
Meanwhile, increasingly capable open-source STT engines offer courts another option, potentially allowing transcription to operate locally without recurring per-hour processing charges. That does not eliminate the costs of implementation, maintenance, or oversight, but it changes the economics considerably.
Still, neither a free transcription engine nor an expensive commercial platform guarantees a successful implementation. The more meaningful question is whether the technology actually reduces the work required to administer proceedings.
If automated transcription helps a clerk locate testimony, prepare notes, or review recordings more efficiently, it can strengthen an already strained workforce. If the same clerk must now troubleshoot audio problems, correct unreliable output, reconcile disconnected files, and manage another application, the promised efficiency may never materialize.
That is ultimately the lesson of speech-to-text adoption. Courts should neither dismiss the technology while waiting for perfection nor rush to embrace it as a substitute for qualified personnel.
The opportunity is real, but realizing it requires investment in the people, audio infrastructure, workflows, and supporting systems that make the technology useful.
Without those investments, even the most advanced transcription engine risks becoming another responsibility assigned to an already overextended workforce.
Cybersecurity shows the cost of disconnected dependencies
Modern courts cannot operate as if cybersecurity is optional. Their systems contain confidential information and support functions that the public depends on. Yet cybersecurity should not be treated as an argument against technology any more than courtroom complexity should be treated as an argument against remote hearings.
The Los Angeles County Superior Court ransomware incident is a sobering example. Following an attack detected July 19, 2024, the court closed all 36 courthouses on July 22 while systems were recovered. Its public statements also described prior investments in cybersecurity that helped the court detect and respond to the intrusion. The lesson is not that modernization caused the attack; it is that continuity, recovery, and security are operational requirements as important as any new software feature. [7]
The federal judiciary’s 2025 annual report likewise describes work on security assessments, asset management, logging, data protection, and threat response. These are ongoing institutional commitments, not one-time purchases. [8]
Figure I illustrates another dimension of the challenge. The number of publicly cataloged software vulnerabilities has grown dramatically over time, reflecting an increasingly complex environment for organizations responsible for securing digital infrastructure.
These figures do not measure attacks against courts specifically, nor do they establish that newer technologies or artificial intelligence are responsible for the increase. What they do illustrate is the scale of the vulnerability-management challenge confronting organizations that depend on interconnected software and hardware.
For courts, the practical implication is straightforward: every new technology investment must also account for the work required to secure, maintain, and recover it.
Figure H. Senior Article III Judges as a Share of Serving Judges, 2001–2025
Percentage of serving Article III judges holding senior status.
Source: CVE Program. Figures represent reported vulnerabilities across software and hardware products generally, not vulnerabilities specific to court systems.
And what happens after an outage? Hearings may need rescheduling. Records may be inaccessible. Attorneys, litigants, and staff must make alternative arrangements. In a court already stretched thin, recovery work competes directly with ordinary court work. Cybersecurity is therefore not a separate technological issue; it is part of the capacity problem.
What would meaningful investment actually look like?
If these pressures interact, then addressing them one at a time may produce disappointing results. A court can hire additional employees while leaving them with cumbersome systems. It can buy excellent software without budgeting for training and integration. It can expand remote access while failing to establish a workable recording and records-management process. None of those decisions is necessarily foolish in isolation. Together, though, they can leave the underlying problem untouched.
Meaningful investment begins with the operation rather than the procurement category. Before adopting a new capability, courts and vendors should examine the work from beginning to end: who prepares the hearing, who controls it, what happens when something fails, who corrects the record, and how information reaches the next person who needs it. The best outcome is not always fewer products. It is fewer unnecessary steps, fewer avoidable handoffs, and less dependence on a single overextended employee.
That also means respecting the variation among courts. In some jurisdictions, the clerk manages recording and conferencing; in others, a dedicated operator or reporter does. A solution that removes work in one setting may simply shift it elsewhere in another. Success should be measured against the local workflow, not just against a vendor feature list.
The responsibility is shared. Legislatures and funding bodies must consider sustainable staffing and operational support, not only one-time capital projects. Court leaders need time and tools to evaluate changes rather than implementing them in perpetual crisis mode. Vendors need to understand how the courtroom works and account for integration, accessibility, security, support, and lifecycle costs. And courts can learn from one another: the same problems are too often being solved independently across jurisdictions.
Conclusion: Invest in capacity, not merely in components
Our starting point was a familiar complaint: courts are being asked to do more with less. The data adds texture to that complaint, and in several places complicates it. National caseloads do not tell one simple story. Salary comparisons do not independently explain recruitment. Federal judgeship trends cannot be applied wholesale to state courts. Technology has demonstrably helped courts continue operating, including during the pandemic.
But those qualifications do not make the underlying question disappear. They make it more important. If a workforce is difficult to replenish, if judicial and administrative capacity cannot adjust easily, and if every modernization project changes the work that employees must perform, then solving each issue independently may never be enough.
Considered individually, none of these datasets proves that American courts are experiencing a single, unified crisis. The compensation figures cannot explain every staffing vacancy. Federal caseload statistics cannot capture the conditions of every courtroom. Judicial population ratios cannot tell us how effectively cases are being managed, and cybersecurity vulnerability counts cannot measure the operational risk facing any particular court.
But considered together, they raise a question worth taking seriously: what happens when institutions are asked to absorb increasing responsibilities while the people, infrastructure, and resources supporting them struggle to adapt at the same pace?
The underinvestment crisis, as we see it, is a problem of alignment: between obligations and staffing, between judicial demands and capacity, between technology and the work it is supposed to support, and between the promise of modernization and the resources required to sustain it.
Perhaps the most useful test for any new investment is also the simplest: does it make administering justice easier for the people responsible for doing so?
We do not claim that this paper establishes a single cause for the challenges facing American courts. We do believe the overlapping trends justify a more connected conversation—among court staff, judges, administrators, policymakers, and technology providers—about what sustainable investment should accomplish.
Courts need more than another system to operate. They need the capacity to operate the systems they already have—and the freedom to choose new ones that genuinely make the work easier.
Sources and further reading
[1] Liberty Recording, “The Underinvestment Crisis in the Courtroom,” October 21, 2026. https://www.libertyrecording.com/p/the-underinvestment-crisis-in-the-courtroom
[2] U.S. Bureau of Labor Statistics, Occupational Employment and Wage Statistics; occupation groups 43-4031 and 23-1012. https://www.bls.gov/oes/ ; see also https://www.bls.gov/ooh/office-and-administrative-support/information-clerks.htm
[3] National Center for State Courts, “Meeting operational demands in a changing environment,” August 20, 2026. https://www.ncsc.org/resources-courts/meeting-operational-demands-changing-environment
[4] Administrative Office of the U.S. Courts, “U.S. District Courts — Judicial Business 2025,” Table 3. https://www.uscourts.gov/data-news/reports/statistical-reports/judicial-business-united-states-courts/judicial-business-2025/us-district-courts-judicial-business-2025
[5] National Center for State Courts, Court Statistics Project. https://www.ncsc.org/our-centers-projects/court-statistics-project
[6] Judicial Conference, “2025 Additional Judgeships Recommendations”; AO, “The Federal Bench — Annual Report 2025.” https://www.uscourts.gov/sites/default/files/2025-03/2025_judicial_conference_recommendations.pdf ; https://www.uscourts.gov/data-news/reports/annual-reports/directors-annual-report/annual-report-2025/federal-bench-annual-report-2025
[7] Superior Court of Los Angeles County, press release July 19, 2024 and General Order July 21, 2024. https://www.lacourt.org/newsmedia/uploads/1420247309641NR07-19-2024SUPERIORCOURTOFLOSANGELESCOUNTYEXPERIENCESRANSOMWAREATTACK-2.pdf ; https://file.lacounty.gov/SDSInter/lac/1164426_NR7212024-COURTCLOSEDTOMORROWASWORKCONTINUESTOGETSYSTEMSBACKONLINEAFTERRANSOMWAREATTACK.pdf
[8] Administrative Office of the U.S. Courts, “Information Systems and Cybersecurity — Annual Report 2025.” https://www.uscourts.gov/data-news/reports/annual-reports/directors-annual-report/annual-report-2025/information-systems-and-cybersecurity-annual-report-2025
[9] Administrative Office of the U.S. Courts, “Annual Report 2025.” https://www.uscourts.gov/data-news/reports/annual-reports/directors-annual-report/annual-report-2025
[10] Research preprint concerning judicial case backlog in India. https://arxiv.org/html/2307.12549v1
[11] U.S. Congress, H.R. 5316, 101st Congress. https://www.congress.gov/bill/101st-congress/house-bill/5316
[12] R Street Institute, “The Right and Wrong Path for New Judgeship Legislation.” https://www.rstreet.org/commentary/the-right-and-wrong-path-for-new-judgeship-legislation/
[13] Administrative Office of the U.S. Courts, “Judiciary Seeks $9.1 Billion in FY 2024 Budget Request,” March 9, 2023. https://www.uscourts.gov/data-news/judiciary-news/2023/03/09/judiciary-seeks-91-billion-fy-2024-budget-request
[14] Research preprint concerning pseudolaw and/or AI in legal contexts. https://arxiv.org/pdf/2503.13556
[15] Research preprint concerning large language models. https://arxiv.org/abs/2409.05746
[16] Duke University Libraries, “It’s 2026: Why Are LLMs Still Hallucinating?” January 5, 2026. https://blogs.library.duke.edu/blog/2026/01/05/its-2026-why-are-llms-still-hallucinating/
[17] Duane Morris, “Stand and E-Deliver: Ransomware v. Courts,” August 2024. https://www.duanemorris.com/articles/stand_and_e-deliver_ransomware_v_courts_0824.html
[18] FBI Internet Crime Complaint Center, “2025 IC3 Annual Report.” https://www.ic3.gov/AnnualReport/Reports/2025_IC3Report.pdf