Dataset

The database as a dataset

Name: The Detained: political prisoners in Russia, Belarus, the Caucasus and Central Asia
Creator: The Detained
License: https://creativecommons.org/licenses/by/4.0/

Every published record, downloadable as CSV or JSON, with a codebook, a stable identifier per person, a fixed-date monthly count and a licence that lets you reuse it. Each record names who designated the person a political prisoner, or states that the assessment is ours.

Download CSV Download JSON API documentation →

The monthly figure

Taken on the first day of every month at 06:00 UTC and never revised. Quote the number with its date. Between snapshots the live database keeps changing as cases develop.

As of 1 October 2026

Published records

3,286

People documented as political prisoners, individual records only.

Re-verified by a person

Records a contributor or editor has checked line by line against the sources.

Under review

173

Held back from publication and from this dataset until an editor confirms them.

Recognised by a national body

1,767

Memorial or the Belarusian joint statement. Documented: 1,099. Reported: 420.

The published figure is not a claim that every person has been independently confirmed by us. The re-verified count is. The distinction is on every record and in the codebook.

Country	Published	Recognised	Documented	Reported
Russia	1,494	520	738	236
Belarus	1,247	1,246	0	1
Georgia	211	0	132	79
Kazakhstan	155	1	117	37
Kyrgyzstan	81	0	44	37
Uzbekistan	95	0	84	11
Tajikistan	74	0	29	45
Turkmenistan	10	0	10	0
Other countries	1	see the CSV

What is in it

Included

Every record with status Published: a person deprived of liberty who meets the definition applied by the criteria.
Individual records only. Multi-defendant cases are published separately at /groups.
Charges, statute references, sentence, custody, dates, sources, and the designation basis for each person.
Summaries, key facts and the rights engaged, in the JSON distribution.

Not included

Records under editorial review. They are counted above and withheld until confirmed.
Inferred demographics. Ethnicity, religion and region of origin are model inferences about individuals and are not released in bulk.
Photographs, which remain under their source licences. See data sources.
Journalist records sourced from the CPJ census, pending permission to republish.

Who says they are a political prisoner

Every row carries a designation tier. Where a national body issues recognition statements under the Guidelines on the Definition of Political Prisoner, the designation is theirs and we cite it. Where none exists, the assessment is ours and the row says so.

Recognised

Listed by a national recognition body applying the Guidelines. Russia: Political Prisoners Support. Memorial. Belarus: the joint statement led by Viasna.

Documented

A human rights organisation documents the case. The Detained assessed it against the Guidelines. This is the usual tier in Georgia, Kazakhstan, Kyrgyzstan and Uzbekistan, where no recognition body exists.

Reported

The case is known from journalism. The Detained assessed it against the Guidelines. Treat these rows as ours, not as a source organisation's designation.

Read the designation policy →

Licence and citation

The dataset is released under Creative Commons Attribution 4.0. Reuse it, build on it, and attribute The Detained. Source records remain under their own terms; when you quote a case, credit the documenting organisation named in its sources column as well.

APA

The Detained. (2026). The Detained: political prisoners in Russia, Belarus, the Caucasus and Central Asia (Version 1.0, snapshot 2026-10-01) [Data set]. https://thedetained.org/data

BibTeX

@misc{thedetained_dataset_20261001,
  title        = {The Detained: political prisoners in Russia, Belarus, the Caucasus and Central Asia},
  author       = {{The Detained}},
  year         = {2026},
  version      = {1.0},
  note         = {Snapshot 2026-10-01. Licensed CC BY 4.0.},
  url          = {https://thedetained.org/data}
}

DOIA DOI is issued when a snapshot is deposited on Zenodo. Until the first deposit, cite the URL and the snapshot date above. Each record page also carries its own citation block with a record version date.

Codebook

Columns in the order they appear in the CSV. Columns marked JSON only are in the JSON distribution, which otherwise has the same fields.

Column	Type	Description
td_id	string	Stable record identifier. Never reused; merged records redirect to the surviving id.
td_ref	string	Human-readable reference derived from td_id, e.g. TD·RU-0042. The Detained's own reference, not an official docket number.
url	string	Canonical public page for the record on thedetained.org.
name_english	string	Name in Latin transliteration as published.
name_native	string	Name in the original script.
country	string	Country of detention (English name).
region	string	Sub-national region where known (ISO 3166-2 style names, see /regions).
gender	string	As reported by the source. Blank when unknown.
age	integer	Age as reported at the time of the source record. Blank when unknown.
occupation	string	Occupation as reported.
arrest_date	date	Date of arrest or first detention, ISO 8601. Blank when unknown.
arrest_location	string	Place of arrest as reported.
current_location	string	Last reported place of custody.
charges	string[]	Charges as reported, pipe-separated in CSV.
criminal_articles	string[]	Statute references as reported (e.g. "Art. 205.2"), pipe-separated in CSV.
case_category	string	Editorial category of the case (e.g. anti-war speech, religious, journalism).
case_basis	string	What the prosecution rests on (Online Speech, Financial Support, Device Evidence, Informant / Secret Witness, Physical Presence, Intercepted Communications, State Fabrication, Unknown). Model-assisted; see methodology.
sentence_years	number	Sentence length in years (fractional for months). Blank when not sentenced or unknown.
sentence_type	string	prison, restriction of freedom, suspended, fine, compulsory treatment, or other.
release_date	date	Reported or scheduled release date where known.
trial_status	string	Stage of proceedings as last reported.
deceased	boolean	true only for deaths confirmed by an editor (see /memorial).
urgency_level	string	Urgency level: how time-sensitive the case appears in the source reporting. critical = acute risk reported (abuse, denied medical care, hunger strike, disappearance); high = active proceedings (pre-trial detention, ongoing trial); medium = no acute risk reported, also the default when reporting is thin; low = reduced exposure (released pending trial, non-custodial sentence, near release). Not a measure of how much a case matters; see methodology #urgency-levels.
human_verified	boolean	true when a contributor or editor has re-verified the record against its sources.
designation_tier	string	RECOGNIZED: a national recognition body applying the Guidelines on the Definition of Political Prisoner lists this person. DOCUMENTED: a human rights organisation documents the case and The Detained assessed it against the Guidelines. REPORTED: the case is known from journalism and The Detained assessed it. See /data/designations.
designation_body	string	The recognition body (RECOGNIZED) or first documenting organisation (DOCUMENTED). Blank for REPORTED.
designation_body_url	string	Website of designation_body.
national_body_exists	boolean	false where no national recognition body applying the Guidelines exists for the country (Georgia, Kazakhstan, Kyrgyzstan, Uzbekistan, Tajikistan, Turkmenistan); such records carry The Detained's assessment.
sources	string[]	Organisations and outlets whose reporting the record is built from, pipe-separated in CSV.
source_url	string	The original source record the entry was created from.
first_seen	datetime	When the record entered the database, ISO 8601 UTC.
last_updated	datetime	When the record last changed, ISO 8601 UTC.
summary_english	text · JSON only	Case summary in English. JSON distribution only.
key_facts	string[] · JSON only	Bulleted facts as published. JSON distribution only.
legal_violations	string[] · JSON only	Rights engaged, as published on the record. JSON distribution only.

Array columns are pipe-separated in the CSV. Dates are ISO 8601. The file starts with a byte-order mark so spreadsheets open Cyrillic names correctly. How the model-assisted columns are produced is described in the methodology.

Versions and stability

Records change as cases develop. Three things stay fixed so a citation can be checked later: the record identifier, the monthly figure, and any deposited snapshot.

td_id never changes and is never reused. When two records are merged, the retired identifier redirects to the surviving one.
Each record page shows a record version date. Cite it: a later discrepancy is then detectable rather than silent.
Corrections are logged on the transparency page. Disputes go to corrections@thedetained.org.
The schema version is 1.0. Adding a column bumps the minor version; renaming or removing one bumps the major version and is announced here.

Snapshot	Published	Re-verified	Under review	Schema
2026-10-01	3,286	91	173	1.0
2026-09-24 (interim)	3,291	89	174	1.0

Working with the data

Researchers

Need a custom cut, a stable snapshot for a paper, or the codebook explained? Write to press@thedetained.org. We can deposit a dated snapshot on request.

Journalists

The press desk has the current monthly figure, boilerplate and turnaround times.

Source organisations

If your organisation's designations appear here and you want the attribution changed, or want to reconcile lists, write to partners@thedetained.org.