Data access

The record is open to researchers, newsrooms, and accountability bodies. Exports carry your filters and every entry’s grade, and we record who takes the data.

Downloads

CSV and JSON exports of the published record, carrying every entry's grade, codes and source count.

Situation report

We ask who you are before an export. Not to gate the data — to be able to trace a figure back if it is ever misrepresented.

Open aggregate data

Province-by-category-by-year incident counts, free to download and reuse under a CC BY 4.0 licence — no sign-in needed. Small cells are suppressed to protect the people involved, and the figures are documented incidents: a lower bound, not a complete total. Do not compare raw counts across provinces or years as if coverage were equal.

Aggregate CSV Aggregate JSON Map GeoJSON Data dictionary

Licence: CC BY 4.0. Please cite the dataset version and your retrieval date.

Researchers: request incident-level microdata

API

Read-only JSON over the published record. No key is needed for the map and summary endpoints; entry-level access is granted to partners.

GET /api/v1/incidents?days=30&province=Kabul

GET /api/map

GET /api/stats

Rate-limited, and versioned — /api/v1 will not change shape under you.

Request API access Methodology →

How grades work

Every row carries the grade the sourcing supports. Filter on it; do not average it.

Confirmed Reviewed and accepted. Two or more independent sources agree on the core facts.

Corroborated Two or more sources report it, but detail differs or review is still open.

Single-source One report only. Published so the claim is on the record, not as an established fact.

Full methodology →

Violation taxonomy

The category codes used in every export, with the treaty article each engages.

Violation taxonomy →

Who uses the record

Organisations that have agreed to be named.

Harvard University Boston University United Nations UN Human Rights Council