A cache of death certificates reportedly obtained from Rosstat’s internal systems contains hundreds of thousands of records linked to Russia’s war against Ukraine. The data cover the period from the start of the full-scale invasion on February 24, 2022, through the end of August 2026. Their disclosure has given researchers a rare opportunity to compare government death-registration records with independent estimates of wartime losses. But the figures emerging from the leak cannot simply be treated as an exact count of Russian military personnel killed.

Where the data came from

On November 1, the hacking groups Silent Crow and Belarusian Cyber Partisans said they had breached Rosstat’s systems and copied approximately 23 terabytes of information. Rosstat acknowledged cyberattacks between October 29 and November 1 but denied that its databases had been compromised or legally protected information leaked. FACTUM.press examined the attack itself in a separate report on the Rosstat breach.

A dataset containing more than 11 million death records from 2021 through 2026 reached several international newsrooms. According to Bloomberg, it includes information about the deceased person’s age, place of residence, date of registration and circumstances of death. This is neither a public Defense Ministry casualty bulletin nor a list of service members compiled from obituaries. It is purportedly an extract from the administrative system used to register deaths. That makes it potentially valuable for identifying losses never mentioned by local officials or the families of those who died.

Three figures from different researchers

Cyber Partisans reported 393,881 deaths linked to military operations and their consequences. Their selection included records carrying specific cause-of-death designations and references to combat. The researchers themselves cautioned that this category might also include civilians and victims of other armed conflicts.

Silent Crow used a broader method. According to an account published by Novaya Gazeta Europe, the group identified 440,982 cases, including not only deaths directly attributed to combat but also 14,160 non-combat deaths of military personnel and 13,671 deaths of people registered in Russia who died in occupied territories or Ukraine without a stated cause. It then added a statistical estimate of another 3,734–9,118 cases among records with no cause of death recorded. That produced a range of 444,716–450,100.

Bloomberg, examining its own sample, identified approximately 392,700 death certificates in which a connection to the war was explicitly indicated by a code or wording such as “military operation,” “shelling” or “drones.” The Guardian published another estimate: 412,576 war-related records. These differences do not necessarily mean the underlying data conflict. Newsrooms and researchers may have used different filters and inclusion criteria. Without a common published methodology and independent replication, the figures cannot be treated as directly comparable.

The claim that “the leak proved 450,000 Russian soldiers were killed” is therefore inaccurate. At present, the evidence points to hundreds of thousands of records that researchers associate with the war, alongside several preliminary estimates apparently drawn from the same dataset.

Why the records appear credible

Bloomberg examined the structure and metadata of the extract and found no signs that it had been fabricated. The newsroom also compared the total number of deaths with Rosstat statistics previously released to the public. The difference was approximately 0.1% for 2021 and 0.2% for 2022. For 2023, the dataset showed about 1% more deaths than the agency’s published figures, including an additional 3,509 deaths among men aged 20–29.

That close correspondence with official totals is a significant indication that the dataset is based on genuine administrative records. Bloomberg nevertheless made an important qualification: it could not cryptographically authenticate the entire database. Matching aggregate figures also cannot rule out errors in individual fields, an incomplete extract or subsequent alterations to records. Meduza and Mediazona said they had begun their own examination of the material.

There is another limitation. A death certificate reflects an administrative classification. The recorded cause of death does not always establish whether someone was a service member, whether they died in combat or precisely how they died. “War-related deaths,” “military personnel killed” and “combat fatalities” are therefore not interchangeable categories.

War deaths enter the statistics months late

One of the dataset’s most consequential features is the delay between a death and its registration. Bloomberg found that the median delay was about six weeks in the second half of 2023 but had exceeded four months by 2025. Of the war-related deaths that occurred in 2024, roughly one in ten was registered more than 14 months later. Most deaths unrelated to the war were registered within about a week.

In August 2026 alone, Bloomberg reported, more than 2,500 war-related deaths from 2025 were added to the database. Records for 2026 therefore cannot be read as a final tally for the year’s first eight months: many deaths during that period may not yet have entered the system. It also means that an increase in deaths registered in a particular month is not necessarily an increase in people who died during that month.

Registration delays matter most when assessing the latest stages of the war. Comparing incomplete 2026 data with completed years without accounting for reporting lags would artificially depress the estimate. Automatically extrapolating earlier registration patterns carries the opposite risk: it may fail to account for changes in fighting intensity, recordkeeping practices or the composition of those killed.

Who is missing from the database

A person listed as missing in action does not necessarily have a death certificate. A service member who later dies of wounds may be registered under a different date and cause. Not everyone killed while serving in units fighting on Russia’s side will necessarily appear in Russia’s death-registration system. Conversely, some civilians who died in areas affected by fighting may fall within the broad categories used by the hackers.

Bloomberg found that, from mid-2024 onward, almost half of the war-related death records lacked a region of residence. Among other deaths, the share with no region recorded was approximately 5%. Regional rankings based on this dataset therefore require particular caution: they may reflect not only real differences in losses but also uneven completion of administrative records.

Even a fully authentic database would not amount to a comprehensive register of all Russian service members killed. It shows, above all, the deaths that were recorded and classified in particular ways.

How the leak compares with independent counts

Mediazona, the BBC Russian Service and volunteers maintain a named list of Russian military personnel killed, compiled from open sources including relatives’ statements, obituaries, official announcements and photographs of graves. By November 6, 2026, the list contained 285,881 verified names. These are individually documented cases, not an estimate of total losses: many deaths never become public.

Meduza and Mediazona use a different approach, examining inheritance-case records and excess mortality among men. Their research estimated that approximately 352,000 Russian service members had been killed by the end of 2025. That estimate comes from a statistical model, not a list of names or cause-of-death codes. It cannot simply be subtracted from a figure extending to August 2026: the periods, sources and methods differ.

The comparison still matters. The order of magnitude suggested by the leak is not wholly disconnected from earlier independent research. But agreement between estimates does not, by itself, validate every record. And where different calculations draw on overlapping sources or similar assumptions, their agreement does not constitute a fully independent cross-check.

What the official statistics conceal

Russia does not publish a current, comprehensive official tally of military personnel killed. The last widely reported public figure from the Defense Ministry was given in September 2022, when Sergei Shoigu said 5,937 had died. Detailed official figures on military losses did not subsequently become part of regular public reporting. Rosstat also stopped publishing detailed cause-of-death statistics that could have helped independent researchers.

The absence of public statistics does not automatically make any leak authentic. But it helps explain why even an incomplete set of internal records became a major international story. Under normal circumstances, the public could compare different government indicators, scrutinize methodologies and debate the consequences of the war using published data. In Russia, much of that scrutiny now takes place outside official institutions.

What can be said today

The purported Rosstat extract contains more than 11 million mortality records. Different analyses identify approximately 393,000–450,000 cases connected to the war or to broader categories of losses. Bloomberg found that aggregate totals closely matched previously published official statistics but could not conclusively authenticate the full dataset. Documented registration delays and the absence of death records for some missing people mean that registered cases are not equivalent to a complete death toll.

For now, the most important outcome of the leak is not a single headline-grabbing number. It is the emergence of a potentially verifiable new source for examining the human cost of the war, the timing of death registrations and gaps in official recordkeeping. Definitive conclusions will depend on independent scrutiny of the documents and transparent explanations of the methods used — precisely what the Russian authorities have so far failed to provide.