DeathsCaused byPoliceWhat the Numbers Say

In the United States, the count the state did not keep was first kept by journalists. In 2015, The Washington Post launched "Fatal Force," a database tracking people shot and killed by on-duty police officers. At the end of one year, it recorded nearly a thousand. That was twice as many as the federal government had ever documented for a single year.

The newspaper maintained this count for a decade, stopping with the year 2024. The other major database, Mapping Police Violence, maintained by the non-profit Campaign Zero, counts all deaths and not just shootings. According to its February 2026 report, police killed at least 1,314 people in 2025, slightly down from the previous year: the first decrease since 2019. The toll remains preliminary, as some deaths are only uncovered weeks or months later.

Nearly all cases are first reported by the local press. The organization is concerned: as newsrooms in small towns shutter, some deaths risk never being counted at all.

The federal government still has no total: the FBI collects these deaths only from law enforcement agencies that volunteer to report them, and publishes only percentages. From 2015 to 2021, according to The Washington Post, it had captured only about a third of fatal shootings.

Vital statistics fare no better. A study published in The Lancet in 2021 compared death certificates against independent databases. Between 1980 and 2018, it estimated 30,800 deaths due to police violence; more than half do not appear as such in official statistics.

The risk is not distributed at random. According to Mapping Police Violence, 93% of people killed in 2025 were men, and more than half were between 25 and 44 years old. Two out of three times, the encounter began with a call: an emergency number, a passerby flagging down an officer, an alarm. One in five showed signs of mental health issues.

Three-quarters were armed, or reported as such, a little over half with a firearm. By the organization's definition, 22% were not, as it counts among the unarmed those holding a replica, a toy, or a phone.

Adjusted for population, the risk of being killed by police is 2.6 times higher for a Black American than for a White American, and higher still for a Native American. The groups most exposed are also the youngest: more than half of Black and Native American victims were under 35 years old, while the majority of White victims were 35 or older. The Lancet study found a gap of the same magnitude over forty years, at equal age.

A rate per resident measures risk, not causation. The disparity is documented across all sources; its explanation remains debated, and these figures are not enough to settle it.

A tally adds up. It does not judge. A death occurring during a police intervention is not, in itself, illegitimate violence: the law authorizes police officers to use force, including a weapon, under specific conditions.

In the United States, the benchmark is a 1989 Supreme Court ruling, Graham v. Connor. The force used is evaluated "from the perspective of a reasonable officer on the scene," rather than with the benefit of hindsight. The Court noted that officers must often make split-second decisions.

In France, the Internal Security Code authorizes police officers to use their firearms "in case of absolute necessity and in a strictly proportionate manner," in specifically enumerated scenarios.

Between the lawful use of force and wrongdoing, only an investigation, and then a judge, can decide. The General Inspectorate of the National Police (IGPN) states this regarding its own count: it "is not intended to determine the legitimacy of the use of force or firearms," because "the legitimate or illegitimate nature of the use of force is in no way determined by the outcome it produces."

Beaumont-sur-Oise, July 19, 2016, on a day of intense heat. Adama Traoré, 24 years old, died about two hours after his arrest by gendarmes, in the courtyard of the Persan barracks. Successive medical evaluations followed without agreement: court-ordered expert reports exonerated the gendarmes, family-commissioned reports concluded asphyxiation, and four Belgian experts concluded in 2021 that a "heatstroke" would "probably" not have been fatal without the arrest. In the absence of a single cause, experts ultimately identified multiple: "multiple factors," none of which, "taken in isolation, was in itself fatal."

The three gendarmes, never placed under formal investigation, were granted a dismissal in August 2023. In May 2024, the Paris Court of Appeal upheld it: their actions had contributed to the death, but remained within the framework of the law. On February 11, 2026, the Court of Cassation made this dismissal final, and the family announced it was taking the case to the European Court of Human Rights. In law, contributing to a death and committing a wrongful act are not the same thing.

Paris, January 3, 2020. Cédric Chouviat, a 42-year-old delivery driver, was stopped on his scooter by four police officers; pinned face down on the ground with his helmet on, he repeated "I'm suffocating" nine times. He died two days later in the hospital. Three police officers were sent for trial before the Paris tribunal correctionnel, the court for lesser offenses, for involuntary manslaughter, while the fourth will not be tried; on October 26, 2026, the trial was postponed to June 21–23, 2027, as the family contends the case belongs before a cour criminelle, which tries more serious crimes. Before judging, it must still be decided whether this death was an accident or a crime.

Philip Stinson, a criminologist at Bowling Green State University, tracks arrested police officers in the United States. Between January 2005 and the end of July 2026, 218 local police officers were charged with murder or manslaughter following an on-duty fatal shooting, during a period when police fatally shot around a thousand people each year. By that date, 78 had been convicted; 107 had been acquitted or cleared through dropped charges, with other cases still pending.

Criminal prosecutions against a police officer following a death remain rare, even if, according to Mapping Police Violence, they have become slightly more frequent since 2019.

The number of people killed did not decline after 2020. Relative to population, it remained stable from 2013 to 2020, rose slightly in 2021, and rose more noticeably from 2022 to 2024; the decrease in 2025 rests on a preliminary tally.

The national registry of the National Police dates back to 2018. That year, it launched an experimental tally of individuals seriously injured or deceased "during the performance of missions of the National Police." A ministerial order of December 14, 2023, made it permanent. The Gendarmerie publishes its own in the annual report of its inspection body, the IGGN.

In 2024, the IGPN recorded 47 deceased individuals during a National Police mission: forty-six men and one woman, no minors. The details show the degree to which this total gathers deaths that bear no resemblance to one another.

Sixteen were killed by an officer's weapon. Eleven died on the road, including nine when a vehicle fled a police stop; among them was a scooter rider uninvolved in the flight, struck by the fleeing vehicle. For the others, the causes were their health condition, suicide, or a fall. Fifteen of these deaths occurred on police premises.

From 17 deaths the first year, the tally reached 47 in 2024. The increase also reflects better recording: police departments are reporting more and more cases, and in December 2024, the Director General of the National Police reminded them all of their obligation to do so.

In 2024, the IGPN investigated fourteen fatal shootings by police officers. In half the cases, the individual was holding a bladed weapon; elsewhere, a firearm, a Molotov cocktail, but also a replica weapon or a mobile phone. In twelve of these investigations, the courts concluded the shooting was legitimate; the other two were still ongoing.

For the Gendarmerie, the IGGN reported 22 deaths in 2024, compared with 13 the previous year. Twelve were killed following the use of weapons. According to IGGN files, seven of these deaths occurred in New Caledonia, mostly during clashes where gendarmes came under gunfire during the 2024 wave of violence.

Independent outlet Basta! maintains its own database of deaths "following law enforcement action," compiled case by case from press reports and official records; like the official tally, it does not judge the legitimacy of the use of force.

In October 2025, an investigation by Politis, conducted with Basta! and the Bondy Blog, recorded at least 560 deaths during law enforcement missions since 2005, the year two teenagers, Zyed Benna and Bouna Traoré, died while trying to escape a police stop. According to this investigation, "the typical profile of the deceased is a non-white young man from a working-class background." Nearly one out of three times, death occurs following a stop or attempted stop, and half of those who died while fleeing were under 22 years old.

For the single year 2024, the investigation counted more than 65, while the two inspectorates combined, police and gendarmerie, recorded 69. For that year, journalists and the administration arrived at the same order of magnitude.

The Ministry of the Interior disputes certain non-profit estimates. In a response to a member of parliament published in December 2024, it criticized "approximate methodologies" and the practice of "conflating the use of physical or armed force with an illegitimate use of force."

The parliamentary mission created after the death of Nahel Merzouk, killed in Nanterre in June 2023, examined a law—that of February 28, 2017—whose responsibility, it writes, is "very often pointed to" in the trend of shootings at moving vehicles.

This law established a common framework for the use of weapons by both police officers and gendarmes. It authorizes shooting when officers cannot otherwise immobilize a vehicle whose driver refuses to stop and whose occupants "are likely to commit, in their flight, attempts on their lives or physical integrity or those of others," subject to absolute necessity and strict proportionality.

In September 2022, a study by Sebastian Roché (CNRS) and two colleagues in Esprit concluded that five times more people were killed by police gunfire targeting moving vehicles between 2017 and 2021 than between 2012 and 2016.

The National Assembly mission declined to determine a cause-and-effect relationship. Its two rapporteurs, one from the presidential majority, the other a Socialist, observed more total shootings following the law, but relative to refusals to comply, the share resulting in gunfire returned as early as 2019 to its pre-2017 level (0.6 to 0.7%).

Since then, these shootings have declined markedly: the National Police reported 138 in 2022 and 87 in 2024. In 2024, none resulted in death. For 2022, two official figures exist: thirteen fatal shootings according to the IGPN, sixteen deaths according to the Ministry of the Interior, in its response to written question No. 549.

The legal framework is under debate once again. On July 7, 2026, the National Assembly passed at first reading a bill establishing a presumption of self-defense for police officers and gendarmes who have used their weapon. This presumption "may, at any time, be rebutted by any contrary evidence." As of the date of this text, the law has not been definitively adopted.

In Nanterre, the question is not yet the guilt of the officer who shot Nahel Merzouk, but the charge itself. In June 2025, two investigating judges referred him to an assize court for murder; in March 2026, the Versailles Court of Appeal dismissed the intent to kill and referred him for violence resulting in death without intent to cause it; in June 2026, the Court of Cassation overturned that ruling for lack of justification. Two jurisdictions answered the same question—did he intend to kill—in opposite ways, and the third sent it back. Reexamined on September 29, 2026, the case awaits a new ruling from the Versailles Court of Appeal, scheduled for November 24, 2026; the officer remains presumed innocent.

In New Caledonia, on May 15, 2024, a young mobile gendarme was killed by armed assailants, in the opening days of the violence. Police officers and gendarmes also die in the line of duty.

In the United States, according to the FBI, 53 police officers were feloniously killed in 2025, including 43 by firearm, compared with 64 the previous year. Accidental deaths, primarily in traffic incidents, are counted separately.

In France, the Minister of the Interior's message for National Police Day in July 2026 noted 39 National Police officers who died on duty or in service since 2021, without distinguishing assaults from accidents, and about 1,100 police officers injured each month. On the Gendarmerie side, the IGGN recorded in 2024 two military personnel who died following an assault and 3,162 gendarmes injured.

In England and Wales, four. That is the number of people shot and killed by police in 2025-2026, according to the IOPC, the independent body that annually publishes deaths occurring during or following contact with the police. Added to this are 15 police-related road traffic fatalities, 27 in or following custody, and 67 apparent suicides following custody—a category absent from the French counts cited here. The watchdog notes that mental health issues, drugs, and alcohol feature frequently in these deaths.

In Germany, official police shooting statistics recorded 16 people shot and killed in 2025. The previous year, with 21 deaths, marked the highest level since the start of national record-keeping. In Canada, no administration maintains a national count; an academic project, Tracking (In)Justice, recorded 745 deaths involving intentional use of force between 2000 and 2023.

Outside wealthy nations, the scale changes again. In Brazil, according to the 20th Brazilian Yearbook of Public Safety, 6,602 people died in police interventions in 2025, the deadliest year in the series. Victims were 97% men, and young Black individuals aged 20 to 29 died 3.5 times more often than White people.

Comparing these countries provides only an order of magnitude, as counting methods differ. But the gap primarily reflects two opposing material realities. In Great Britain (England, Wales, and Scotland), police officers patrol without firearms: only specially trained units are armed. In the United States, where more civilian firearms circulate than residents, every policing doctrine operates on the assumption that a traffic stop or night call could escalate into an armed confrontation. Adjusted for population, American police fatally shoot about eight times more people than French forces, more than fifteen times more than German police, and roughly fifty times more than police in England and Wales.

Official sources depend on what agencies report, and journalists' databases on what the press reported. Where the local press remains silent, and where an agency is slow to report, a death can leave no line at all.

Injuries remain poorly tracked, and the regulations governing the French registry show why: an injury is entered only if it leads to at least nine days of total incapacity for work and a judicial investigation was opened. In 2024, the IGPN recorded 1,585 reported uses of defense ball launchers (LBDs), and only a single person injured by this weapon, down from 20 in 2023. What the registry does not capture does not exist in the statistics.

An injured person enters no death count, but shows who writes the first draft. Paris, November 21, 2020: music producer Michel Zecler was beaten during an arrest inside his studio, a scene captured by a surveillance camera, and survived. According to the Defender of Rights, the incident report drafted by the officers "does not faithfully reflect" the intervention. Three police officers are scheduled to be tried from November 9 to 19, 2026, before the Paris departmental criminal court, for aggravated willful violence and forgery in public records; the racist nature of the assault was dismissed by investigating magistrates.

The Washington Post database dates to 2015, the French official registry to 2018. Before them, a portion of these deaths appeared in no statistics at all. Minneapolis had its verdict, Nanterre its parliamentary mission; the other dead appear, at best, in a count.

In France as in the United States, these numbers exist only because someone decided to count them.