What this site holds back

mappingpoliceviolence.us states: “No part of this site, Mapping Police Violence, may be reproduced in whole or in part in any manner without the permission of the copyright owner.” This site’s records hold no permission document from the copyright owner. That is a gap in provenance, and it is stated here as one.

The site’s current wording is left in place while the written scope is confirmed; Source inventory says the same. Older wording describing this data as used “under written permission” remains in /downloads, /llms.txt, /llms-full.txt and the agency API, and in two internal files: a comment in the query layer and the extraction script for this data. Those were not revised when the other pages were, and they are listed here so the difference is visible.

No licence, no public-domain dedication and no re-use statement was found on TCOLE’s racial-profiling page, its website policies, its privacy and security policy, its link policy or its public-information page, and the workbooks carry no rights metadata. The only rights statement on the site is a footer: “Copyright © 2023 Texas Commission on Law Enforcement”. A copyright notice asserts ownership; it grants nothing.

TCOLE publishes the workbooks for download, for the public, with no stated condition. That is publication, not a licence. As of 12 October 2026 TCOLE had not been asked what re-use it permits. The evidence is in data/licences/tcole-racial-profiling-2026-10-12.md.

Absence is not zero and cannot be made into zero. The FBI’s per-agency API answers HTTP 200 with every count 0 for an ORI that does not exist (probed with ZZ9999999), so an API zero is not evidence that an agency reported none. Nothing this site holds distinguishes “reported zero” from “did not report”.

The file is silent on most of the register. In 2025, 3,058 of the 19,645 agencies in the register have at least one incident in it and 16,587 have none. Across every agency in the register and every year in the file, 617,524 of 687,575 agency-years (89.8%) are silent. A silent agency-year may be zero incidents, non-reporting or non-participation; the data cannot say which, so no page says any of them.

New York City Police Department, 2023: 612 in the bulk file, 624 in the API. 2024: 657 in the bulk file, 648 in the API. Two smaller agencies differ as well: Fort Lauderdale 2023 (14 in the bulk file, 6 in the API) and Manchester, New Hampshire 2025 (3 and 2). The bulk figures are read from this site’s stored rows; the API figures are from probes made on 12 October 2026, and the API is not stored.

The stored rows match the bulk file row for row, so this is not a parsing error here. The API reports a last refresh of 15 September 2026 and the bulk file was last modified 14 August 2026, which is consistent with the two routes carrying different revisions. That is a hypothesis; it is unproven. These four were found in a sample of agency-years, not a full comparison, so others may exist.

POST’s own footnote says: “Last Employing Agency” is not necessarily the reporting agency or the agency where the serious misconduct occurred. A count on a department’s page would assert what the publisher disclaims.

POST also states that it “is aware that there are issues affecting the accuracy of a peace officer’s certification status as listed on this page”, and the endpoint is not deterministic: one officer alternated between MADERA CO SO and MADERA CO DA across 12 identical requests.