They're all cops. The file, the archive, and how the work is built.
This hub routes to three places. The structured dataset is the quantitative core: 63,039 canonical incidents merged and cross-validated across five sources, spanning 1930 through 2026 and covering 32 countries. The news archive is the volume argument: four years of community-curated reporting on misconduct broadly. The methodology page is the trust-building: dataset adapters, dedup strategy, libel rails, and what’s coming.
The scope runs wider than killings now. Mapping Police Violence, the Washington Post’s Fatal Force, and Fatal Encounters document police-involved deaths. The Invisible Institute’s Citizens Police Data Project adds 24,307 Chicago PD sustained complaints covering excessive force, civil rights violations, corruption, sexual misconduct, and everything else the department’s own internal investigators confirmed. The Discord archive extractor pulls another 3,642 misconduct cases from four years of community-curated news, including DV, bribery, evidence tampering, and officer extremism that the killing datasets don’t touch.
From an anarchist analysis, the distinction between a beat officer, a deputy, a corrections officer, a federal agent, a judge, and a prosecutor is organizational not structural. They share a function: the coercive arm of the state. They’re all cops.
63,039 documented incidents from 1930 to 2026. Five sources merged and deduped: Mapping Police Violence, WaPo Fatal Force, Fatal Encounters, Chicago's CPDP, and the Discord archive classifier. 12,904 rows are cross-validated across multiple datasets. Filter by year, source, status, armed status, incident type, US state, or country (32 represented).
Four-year archive of 12,686 community-curated news links documenting misconduct broadly. Top 25 outlet breakdown. Month-by-month posting volume. 750-row sample across all years, year-chip filtered.
How this is built. Per-dataset adapters, the cross-dataset dedup algorithm, libel and ethics rails, country-tagging logic, and provenance links for every source.
The whole file in one shot. 63,039 canonical rows as CSV or JSON, with a schema reference, dedup notes, and a CC BY 4.0 license. Free to use with attribution ... research, journalism, visualization, whatever you build with it.
How many more do you need.
