The project
Document, cross-reference, transmit
This map gathers, category by category, the war crimes committed by Russian forces since the beginning of the aggression against Ukraine in 2014. Every crime is tied to a place, a date and a legal qualification, so the corpus can be queried across dimensions: for instance the destruction of hospitals in a given region between two dates.
- Categories
- 12
- Crimes
- 2,550
- Places
- 786
- Victims
- 967,438
Methodology
Scope
The corpus records serious international crimes committed by Russian forces in Ukraine from 24 February 2014 onwards: the annexation of Crimea, the conflict in Donbas, and the full-scale invasion launched on 24 February 2022. It covers attacks on civilians, protected objects, deportations, environmental damage and other violations of international humanitarian law attributable to the Russian Federation or its proxies. It does not document Ukrainian military operations or losses.
Sources
Each record is sourced from a published report by an independent organisation. The main sources are: UN Human Rights Monitoring Mission in Ukraine (HRMMU/OHCHR), the Commission of Inquiry on Ukraine, the OSCE, Human Rights Watch, Amnesty International, Bellingcat, The Kyiv Independent war crimes unit, the Center for Civil Liberties / Tribunal for Putin, UNESCO, WHO and the Office of the Prosecutor General of Ukraine. We do not publish anonymous allegations; every incident carries one or more URLs.
Verification levels
Three levels are used. Documented: corroborated by at least two independent sources or by a UN/international inquiry. Reported: documented by one recognised organisation or official investigation. Alleged: mentioned in credible media or official statements, awaiting further corroboration.
Collection and update rhythm
Curated records are entered by researchers. The Bellingcat civilian-harm dataset is imported automatically every week. OHCHR verified casualty figures are refreshed monthly, with historical periodic reports backfilled to February 2022. Front-line geometry comes from DeepState and is refreshed every three days. Deportation flows and protected-site damage are updated as new reports appear.
Classification
Each incident is classified under one crime type (for example attack on a civilian object, torture, deportation, environmental crime). The type carries a short legal qualification referring to the Rome Statute and the Geneva Conventions. The classification is ours and is intended for research, not as a judicial determination.
Geography and dates
Places are geocoded to an oblast and to x/y coordinates on the SVG map. When an incident concerns an entire region or the whole country, it is attached to a 'nationwide' or 'occupied territories' virtual place rather than a city. Dates are recorded with one of three precisions: exact date, month known or year known.
Casualty figures
Two indicators are kept strictly separate. The map counts the killed, injured and affected individuals recorded in each incident file. The OHCHR/HCDH panel shows UN-verified civilian casualties. The two series are never added together: the file-level counts are partial and event-based, while the UN figures are minimum verified totals.
Traceability of UN figures
Every UN figure displayed on the map, in the timeline and in the panels is attached to the published OHCHR report it comes from, exactly like the sources of an individual case file. Hovering an oblast or a timeline bar names the report; the detail panels link straight to it. Monthly updates take precedence over periodic reports, which are pro-rated across the months they cover and flagged as such.
Limits and intended use
The database is partial and non-exhaustive. It reflects what has been documented and published, not the full scale of violations. It is offered to journalists, researchers and legal teams as a starting point; any citation should link to the original source. The project does not replace a court judgment.
AI and automation
Some repetitive tasks use language models: extracting structured casualty figures from OHCHR monthly PDFs, suggesting incident records from press releases, and producing the cross-analysis texts on the AI analysis page. Every machine-suggested record is reviewed against the original source before publication; the analysis outputs are clearly labelled as AI-generated and must be checked against the linked sources.
Corrections and contributions
Researchers, journalists and witnesses can suggest corrections or new records via the Contact page. Every submission is reviewed before it is added to the map. When two sources describe the same event, the records are merged and all source links are preserved.
How to read the map
- Stackable layers — point crimes, density by oblast, occupied territories by period, protected sites (hospitals, schools, heritage) and deportation flows can be displayed at once, with an opacity control to compare them.
- Cross-filtering — crime type, period and region combine. The counter of each category recalculates according to the other active filters.
- Previews — hovering a point or a category shows a summary; clicking opens a detailed record with chronology, sources and testimonies.
- Sharing — the filter state is written into the page address: a cross-filtered search can be shared as is.
Testimonies
0 testimonies are attached to the documented places. In the production version, each one links to a video interview and to the crimes it describes.
Data and sources
Every record comes from a published source: Center for Civil Liberties / Tribunal for Putin, The Kyiv Independent war crimes unit, UN human rights bodies and the Commission of Inquiry, OSCE, Human Rights Watch, Amnesty International, UNESCO, WHO and the Ukrainian Prosecutor General's Office. Each crime carries its links, a verification level and a date precision; the coverage is partial and grows batch by batch.
Open the map