Use of Force in 2025: What Federal Data Tells Us

An analysis of the FBI National Use-of-Force Data Collection, its coverage, its gaps, and what the visible patterns mean for law enforcement training.

When a police executive wants to answer a basic question, how often do officers use serious force, and under what circumstances, the honest answer in 2025 is still: we have a partial picture. The United States has never operated a complete, mandatory national count of police use of force. The FBI's National Use-of-Force Data Collection (NUSF) is the most ambitious federal attempt to build one, and after six years of operation it offers real insight. It also illustrates, clearly, the limits of voluntary reporting.

This article reviews the state of the NUSF program as of 2025: what it measures, how much of the country it covers, what patterns are visible in the aggregated data, and what those patterns suggest for agencies designing their training programs.

What the NUSF program actually measures

The National Use-of-Force Data Collection launched on January 1, 2019, following years of development work led jointly by the FBI's Uniform Crime Reporting (UCR) Program and the Bureau of Justice Statistics.¹ Participation is voluntary, and it is open to all federal, state, local, and tribal law enforcement and investigative agencies.²

The collection has a deliberately narrow definition. Agencies are asked to report only three categories of event: any use of force that results in the death of a person, any use of force that results in serious bodily injury, and any incident in which an officer discharges a firearm at or in the direction of a person.² It does not capture the far larger universe of lower-level force, such as a controlled handcuffing, a brief use of a firm grip, or the display of a weapon that is never discharged. This is an important framing point: NUSF is a serious-force collection, not a general use-of-force collection.

The data is published as aggregate national statistics on the FBI's Crime Data Explorer (CDE).³ The FBI is explicit that the collection "offers big-picture insights, rather than information on specific incidents, and does not assess or report whether officers followed their department's policy or acted lawfully."⁴ In other words, NUSF is a descriptive instrument. It tells us what happened, in broad strokes. It is not a measure of whether what happened was justified.

The participation story

The single most important fact about NUSF is also the most frequently misunderstood: as of 2025, the FBI has never released a national count of use-of-force incidents from the collection. The reason is a self-imposed data-quality rule. The FBI will publicly release incident counts only when participating agencies cover at least 80 percent of the sworn law enforcement population.⁴

Coverage has climbed steadily but has hovered just below that line. Reported coverage rose from roughly 46 percent of the law enforcement population in the program's first year (2019) to 80 percent in 2024, before slipping back to 78 percent in 2025.⁵ In the 2025 reporting year, 12,035 of the nation's 19,277 law enforcement agencies submitted data, including 94 federal agencies, together covering 78 percent of the sworn population.⁶

Two structural points are worth noting. First, in January 2024 the FBI changed how it measures participation, moving to a rolling 12-month span rather than a calendar-year count that reset to zero each January.⁵ This produces a steadier, more honest participation figure and reduces the artificial early-year dips that the old method created. Second, the gap between the 19,277 agencies that exist and the 12,035 that report is not random. Larger agencies are generally better resourced to participate, so the population-coverage figure (78 percent) is higher than the raw agency-count figure (about 62 percent of agencies). The missing agencies skew small and rural, a point that matters for any reader trying to generalize from the data.

The practical consequence: the FBI has published the distribution of use-of-force characteristics, the percentages and patterns, but not the volume. We can say what a typical reported incident looks like. We cannot yet say, from NUSF alone, how many incidents occur nationally each year.

What the visible patterns show

Even without a national count, the aggregated distributions are informative. The figures below reflect FBI NUSF data as compiled and reported for the 2021 through 2025 period.⁶

Reason for the initial contact. The largest share of reported serious-force incidents began with officers responding to unlawful or suspicious activity, about 54.3 percent. Traffic stops accounted for roughly 12.9 percent, and routine patrol activity for about 7.0 percent, with the remainder distributed across other or unknown contact reasons.⁶ The headline takeaway is that most serious-force incidents in the data did not begin as a routine, low-information encounter. They began with officers responding to a reported problem. That does not make the encounters predictable, but it does mean a meaningful share of them involve a window of time before contact in which officers can plan an approach.

Subject behavior. The most commonly reported forms of subject resistance or threat were failure to comply with verbal commands, attempting to flee, displaying a weapon, using a firearm, and resisting arrest or handcuffing.⁶ The presence of "failure to comply with verbal commands" at the top of the list is notable. It is not, by itself, evidence of a communication failure on the officer's part. But it does indicate that a large fraction of serious-force incidents pass through a verbal phase before force is used, which is precisely the phase that communication and de-escalation training is designed to address.

Type of force. Across the reported incidents, firearms were the most frequently cited force method, followed by hands, fists, and feet, then conducted energy devices (tasers), police canines, and impact projectiles.⁶ Firearms ranking first is a direct artifact of the collection's definition: because NUSF captures every firearm discharge at a person regardless of injury, and only the most serious other force, firearm use is structurally over-represented relative to its share of all police force. This is a good example of why the collection's scope must be understood before its numbers are quoted.

Outcomes. Within the reported incidents for the 2021 through 2025 window, roughly 30 percent involved a death, about 60 percent involved serious bodily injury without death, and roughly 10 percent involved a firearm discharged at a person without resulting injury.⁶

Frequency. In each month of 2025, only about 1 percent of participating agencies reported at least one qualifying incident.⁶ Serious use of force, as NUSF defines it, is a rare event at the individual-agency level. For a single mid-sized department, a qualifying incident may occur only a handful of times per year or less. This rarity is itself a training design challenge: officers cannot build proficiency through repetition on the job, which is one of the central arguments for scenario-based rehearsal.

What the data cannot tell us

Intellectual honesty about NUSF requires stating its limits plainly. A working chief reading this should be able to anticipate every objection a skeptic would raise.

First, coverage is incomplete and non-random. At 78 percent population coverage, roughly one in five officers works for an agency not represented in the data, and those agencies are disproportionately small and rural.⁶ Any national generalization carries that caveat.

Second, the collection captures only serious force. It is silent on the much larger volume of lower-level force, which means it cannot be used to study escalation patterns from their earliest stages or to measure how often de-escalation succeeded and force was avoided entirely. Avoided-force events, the goal of de-escalation training, are invisible to NUSF by design.

Third, NUSF does not adjudicate. It records that force occurred, not whether it was lawful, within policy, or avoidable.⁴ No reader should treat a NUSF figure as a count of misconduct, and no reader should treat it as a count of justified action either.

Fourth, the absence of incident counts limits trend analysis. Because the FBI has not crossed the 80 percent release threshold consistently, year-over-year volume comparisons are not yet possible from NUSF. For counts of fatal incidents specifically, researchers continue to rely on supplementary sources such as the CDC's fatal-injury data and non-governmental databases, each with its own methodology.

Fifth, correlation is not causation. The data shows, for example, that many incidents began with a verbal-noncompliance phase. It does not show that better verbal tactics would have changed the outcome in any specific case. The patterns identify where to look for training opportunities. They do not, on their own, prove what training would have prevented.

What this means for your agency's training

The federal data, read carefully, supports a small number of durable conclusions for training design.

The data establishes that serious force is rare at the agency level but consequential when it occurs. That combination, low frequency and high stakes, is the textbook case for scenario-based training rather than reliance on field experience. Officers will not encounter enough qualifying events to build judgment through repetition, so the repetition has to be manufactured in a training environment.

The data establishes that a large share of incidents pass through an identifiable pre-contact and verbal phase. Most begin as a response to a reported problem, giving officers planning time, and many involve a verbal-noncompliance stage before force. Training that rehearses approach planning, threat assessment, and structured communication is therefore addressing the phases where the data says the encounters actually unfold.

The data establishes that traffic stops, while a minority of serious-force incidents, are a meaningful and recurring category, around one in eight. That is enough to justify dedicated traffic-stop training as a distinct module rather than a footnote.

Finally, the data establishes that the public record is incomplete, which is itself an argument for agencies to maintain rigorous internal use-of-force documentation. Agencies that track their own lower-level force and de-escalation outcomes can answer questions about their own performance that NUSF will never answer for them.

CodeBlu's training model is built around scenario rehearsal of exactly these phases: the pre-contact plan, the verbal phase, and the decision point. The federal data does not prove any single program's effectiveness, and this article does not claim that it does. What the data does is identify the encounter stages where serious force concentrates, and a training program is well designed to the extent that it puts officer practice time into those stages.

Recommended CodeBlu scenarios this article supports

1. Response-to-suspicious-activity approach planning: rehearsing the pre-contact phase for the single largest category of serious-force incidents, where officers typically have some planning time.
2. Verbal-noncompliance decision points: structured practice for the encounter phase that precedes a large share of serious-force events.
3. Traffic stop threat assessment and communication: a dedicated module reflecting the recurring traffic-stop share of the data.
4. Low-frequency, high-stakes event rehearsal: repeated exposure to firearm-decision scenarios that officers cannot accumulate through field experience.
5. Use-of-force documentation and debrief: building the internal-data habit that compensates for the gaps in the national record.

---

Footnotes

1. Bureau of Justice Statistics and FBI, "National Data Collection on Police Use of Force." https://bjs.ojp.gov/content/pub/pdf/ndcopuof.pdf ↩

2. FBI, "National Use-of-Force Data Collection." https://www.fbi.gov/how-we-can-help-you/more-fbi-services-and-information/ucr/use-of-force (accessed May 2026). ↩ ↩²

3. FBI Crime Data Explorer. https://cde.ucr.cjis.gov/ (accessed May 2026). ↩

4. FBI, "FBI Releases Use-of-Force Data Update," 2025. https://www.fbi.gov/news/press-releases/fbi-releases-use-of-force-data-update ↩ ↩² ↩³

5. FBI, "FBI Releases 2024 Quarterly Crime Report and Use-of-Force Data Update," 2024. https://www.fbi.gov/news/press-releases/fbi-releases-2024-quarterly-crime-report-and-use-of-force-data-update ↩ ↩²

6. USAFacts, "What FBI data says about law enforcement use-of-force," 2025, summarizing FBI National Use-of-Force Data Collection figures for the 2021 through 2025 reporting period. https://usafacts.org/articles/what-the-data-says-about-law-enforcement-use-of-force/ ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸