Ukraine war casualty assessments – The Leak That Arrived Right on Cue
Wed 2:57 pm +00:00, 11 Nov 2026
Source: https://sonar21.com/the-leak-that-arrived-right-on-cue/
In the intelligence business, timing is everything. On February 5, 2026, Finland’s National Defence University, which is hardly a Kremlin front, posted a paper by Alexander Grable titled “Irretrievable Losses: Casualties in the Russo-Ukraine War (2022–present).” Grable concluded that Ukraine’s dead far outnumber Russia’s and that Kyiv is losing the war of attrition. In polite Western company, that is heresy.
The paper was presented at the university’s Russia Seminar a week later. It was folded into the seminar’s proceedings volume this fall and began circulating widely in late October. Then, right on cue, the cavalry arrived. Two hacker groups announced they had breached Rosstat, Russia’s state statistics agency, and come away with numbers showing between 393,881 and 450,100 Russian dead. The Moscow Times and Euronews ran the story on November 8. Narrative restored. Everyone can relax.
I don’t relax that easily. I spent enough years at the CIA to know what a planted story smells like. This one smells. But before I get to the leak, let me get to something the casualty accountants keep leaving out of their spreadsheets: who has been shooting more.
The math nobody wants to do
Weapons cause casualties, so a casualty estimate that ignores the weapons is guesswork. In this war, Russia has held a lopsided edge in nearly every category that kills soldiers: artillery, mortars, missiles and FAB glide bombs. For most of the war it also led in drones.
Start with artillery. Estonia’s foreign intelligence service, no friend of Moscow, estimates that Russia produced or refurbished about 400,000 rounds in 2022 and 4.5 million in 2024, an elevenfold increase. That same year the United States and Europe together managed about 1.2 million. Then add North Korea: leaked South Korean intelligence, reported in the press last year, put Pyongyang’s deliveries at 12 million rounds of 152mm ammunition. By my count, that adds up to roughly 24 million shells on the Russian side since 2022, against about 7 million for Ukraine. That is better than three to one.
The glide bombs and missiles tell the same story. Ukraine has nothing remotely comparable to Russia’s FAB-500s and FAB-1500s fitted with wing kits, dropped by the thousands on Ukrainian positions from beyond the reach of air defenses.
Timing matters here too. Artillery was the main killer at the start of the war, causing roughly 80 percent of casualties, and that is exactly when Russia’s lead in shells was widest. Drones did not overtake artillery until 2024. The widely quoted estimate that drones cause 70 to 80 percent of battlefield casualties dates from 2025, not from 2022. Kyiv now insists it outproduces Russia in FPV drones. Maybe so. But even a Ukrainian official has conceded that Ukraine was far from parity with Russia in strike drones through the first half of 2025.
The drone race is also being fought at the factory. Russia has repeatedly hit Ukrainian drone production: assembly sites around Kyiv for the Liutyi long-range drone, industrial zones in Kharkiv and Dnipro that make guidance and communications parts, and in October a plant belonging to the manufacturer General Chereshnya. Ukraine has hit back, striking the Atlant Aero drone plant in Taganrog three times between January and April and repeatedly targeting the Alabuga zone in Tatarstan. Yet Russian output keeps climbing. By Ukraine’s own intelligence estimates, Russian drone production in April was up 117 percent from a year earlier, and Moscow plans to build 7.3 million FPV drones this year. Kyiv’s raids have not bent that curve.
The standard Western reply is that attackers lose more men than defenders, and Russia has been attacking. That rule of thumb assumes both sides can shoot back at roughly the same rate. When one side fires three or four shells for every one it receives, the rule goes out the window. Defending under that kind of fire is how armies get ground down.
What Grable actually found
Grable worked from amputations, which is a clever approach because amputees are harder to hide than the dead. He applied historical ratios of deaths to amputations to the amputation counts for each side. For Ukraine, 93,566 to 120,000 amputations yield an estimate of 403,269 to 517,200 dead by November 2024. He also cites body exchanges in which Ukraine got back about 12.5 of its dead for every Russian returned.
Grable also puts Russian dead at about 35,264 by mid-2024. I’ll be candid: that figure is too low. It sits below the number of Russian dead that Mediazona and BBC Russian had already identified by name at that point. His Vietnam-derived multiplier for Russia doesn’t hold up, and his critics are right to hammer it.
But notice what the critics do next. They take one weak number and use it to throw out the whole paper, including the half that matches the firepower. If Russia has been firing three or four times as much ordnance, Ukrainian dead running several times higher than Russian dead is not a wild claim. It is what the physics predicts. Grable’s Ukrainian estimate fits the weight of fire. That is the part that had to be buried.
Consider the source
Now the leak. Look at who delivered it. One group is Silent Crow, which claims about 450,100 Russian dead. The other is the Belarus Cyber Partisans, which claims 393,881. Neither is a neutral archivist. Both are openly at war with Moscow and Minsk, and they say so with pride. We are asked to believe that these partisans pulled 23 terabytes out of Rosstat, sorted it and produced clean totals, all within days. Then Western newsrooms printed the result as “official data.”
Official according to whom? Meduza and Mediazona, which have no love for the Kremlin, say they are still checking whether the data is genuine. Nobody outside the hackers themselves has verified the totals.
The Moscow Times reports that it checked 50 names from the Mediazona list of named Russian dead and found every one in the leaked files. That is supposed to settle it. It settles nothing. The Mediazona list is public. Anyone building a forgery would start by seeding it with names a fact-checker can find. Matching names proves the forger did his homework. It does not prove the totals.
How an influence operation is built
The defenders of the leak have one more argument: the timing is too tight. Grable’s paper went viral on October 26 and 27, and the hacks are dated October 29 and November 1. Nobody, they say, could build a 23-terabyte operation in a week.
That argument assumes the people who counter inconvenient research learn about it from Substack. They don’t. The paper sat in a Finnish government repository from February 5. Anyone working Russia for a Western service, and certainly anyone in the information-warfare shops, had it in hand within days. That gave them eight months, not one week. Public attention is when an operation launches, not when the planning starts.
I worked on this kind of thing at the CIA. My assessment is that this leak has the classic hallmarks of an influence operation. There is a source that cannot be checked, delivered through a cutout with a cause. The material is laundered through outlets that readers trust. It drops just as an unwelcome analysis starts to gain traction. And it comes with a big round number that ends the debate before anyone asks about the weapons. I can’t prove who wrote the script. I have seen enough scripts to recognize one.
The question they’d rather you not ask
Here is the test I’d put to anyone waving the Rosstat numbers. Explain how an army that has been outgunned three or four to one in artillery, with no answer to glide bombs and, until recently, outmatched in drones, somehow lost fewer men than the side doing most of the shooting. If you can’t answer that, a spreadsheet from a hacker collective won’t answer it for you.
Grable got the Russian number wrong. He got the big picture right, and the big picture is what the leak was built to bury. Ukraine is bleeding faster than Russia because Russia has more guns, more shells, more bombs and, for most of this war, more drones. No amount of stolen data changes the ballistics.
