Data snapshot: October 4, 2026, 20:17 UTC. Atlas source data as of 15:21:09 UTC. The country score covers a rolling 30-day OONI measurement window; the July evidence record was read separately at 20:17 UTC.
Russia's row in the Voidly Atlas country index showed 45,085 OONI samples and a score of 36 when we read it on October 4. Atlas put that measured anomaly signal in its “medium” band. A separate July 1 evidence record contained 16 links to OONI observations concerning messaging and circumvention tools. The large rolling sample and the smaller dated trail describe different things. Neither is a count of people blocked or a measure of every Russian network.
A large denominator still has boundaries
Atlas derives its score from the anomaly rate in the OONI samples available for a rolling 30-day period, applies a small-sample adjustment and rounds the result. The published method and country row put the score beside its 45,085-sample denominator and source-data timestamp. A score of 36 is not a claim that 36% of residents lost access. It is a summary of tests that entered this particular calculation.
That sample count is substantial compared with many other country rows in the same index. It does not, by itself, make the sample nationally representative. OONI explains that its measurements depend on who runs Probe, which networks they use, which tests they choose and when they run them. A service can work on one network and fail on another. A countrywide number can hide both differences.
The response also lists services associated with the country row. That list is a guide to questions worth testing, not a finding that every listed service was unavailable across Russia throughout the 30-day period. OONI calls a test “anomalous” when it shows potential interference; anomalies can have other explanations. To support a service-specific statement, a researcher must inspect the same test across networks and time, including successful results and the errors inside anomalous ones.
Read the July record at its own scale
The Russia evidence response lists 16 OONI entries and counts one source family, with low corroboration. The linked entries cover four test families—Signal, WhatsApp, Telegram and Tor—at four listed observation times on July 1. A reader can follow the Signal, WhatsApp, Telegram and Tor searches attached to the record.
The number 16 counts linked entries, not independent monitoring organizations or affected networks. All four test families come from OONI in this response. The links let a reader test a narrower proposition—for example, whether Telegram anomalies repeated on a named network at a given time. They cannot alone establish intent, a single cause across services or nationwide impact.
The July trail should not be used as if it were the source of the October rolling score. The record and index have different windows and purposes. When citing a current score, read the index's dataAsOf field. Its generated_at_utc field records when the API served the response, and can change even when the underlying measurement set has not.
What a network-level investigation looks like
OONI's March 2026 Telegram investigation shows the extra work behind a stronger conclusion. Its chart covered Telegram tests on 547 Russian autonomous systems from February 28 to March 30. OONI saw anomalies rise around March 15 and spike from March 20. It inspected connection resets and timeouts and named networks with strong signals, including Rostelecom (AS12389), Vimpelcom (AS8402), MTS (AS8359) and Central Telegraph (AS8615).
The same report states an important limit: measurement coverage varied by network, and the Telegram block was not visible on every tested network. The investigation supports a finding about many tested ISPs during that March 2026 period. It does not make the July evidence record a March report, or prove that every Russian network had the same result in October. The date, tested service and network breakdown must travel with the conclusion.
This is why the October total of 45,085 samples is useful but insufficient on its own. To study a reported block, inspect the service's failures, the proportion of successful tests, the networks reporting each result and changes over time. OONI also warns that app infrastructure changes and temporary failures can create false positives. Repetition and consistent failure modes make a signal more persuasive; a national aggregate is the start of that analysis.
A defensible reading
For this snapshot, Atlas reported 45,085 OONI samples, a score of 36 and a medium anomaly band for Russia. Its dated July record exposed 16 OONI links across four tests, with one counted source family. OONI's March Telegram report provides a separate, deeper example of how to interpret a particular app on particular networks. Those statements can each be checked at their index, evidence record and upstream study.
The next question should be specific: which service, on which networks, over which dates, with what errors and successful controls? Keep the index denominator visible while answering it. That produces a claim a reader can inspect without turning a measured signal into an unsupported statement about every person in the country.