Moscow did not sound its air-raid sirens. The polls were open.

Begin with the fact that is not in dispute in the report in front of me, because the argument has to be built on it and not on my opinion of the people involved. On the night of 19–20 September, as voting in Russia’s State Duma elections finished, the Moscow region was struck. Russian officials acknowledged strikes across 17 Moscow municipalities as the voting concluded.¹ Moscow’s mayor, Sergei Sobyanin, claimed air defences had downed more than 1,600 Ukrainian drones since 19 September, 450 of them on the approaches to the capital.¹ The Institute for the Study of War describes the barrage as the largest combined missile and drone strike package Ukraine has launched against strategic targets in the Moscow region.¹ That is the shape of the superlative, and I will keep it that narrow.

Now the claim this column is about, and I want its status kept visible the whole way through. According to ISW, authorities in the Moscow region withheld the air-raid sirens and the SMS warnings.¹ They restricted mobile service, which ISW calls a routine countermeasure, and did so without major public announcements.¹ They did not activate the alert sirens or send the SMS notifications that residents had received during far smaller attacks.¹

That last clause is the entire evidentiary weight of the story, and it is worth slowing down for. The claim is not that Russia has no warning system. It is that the system exists, that it has been used for less, and that on the largest night it was not used.¹ A capability withheld is a decision. A capability absent is only a gap. ISW’s assessment is that this was the former, and it reaches that conclusion by comparison with the state’s own past conduct — which is, in my view, the right way to reach it.

I will separate these as plainly as I can, because conflating them is how a strong story becomes an unusable one.

Shown, if ISW’s account is accurate: sirens and SMS alerts went out during significantly smaller previous attacks and did not go out during this one.¹ Mobile service was restricted without major public announcement.¹ Residents of the Moscow region voiced anger about the silence, some of them on the Telegram channel of the regional governor, Andrey Vorobyov — until that channel disabled comments.¹ A prominent Kremlin-affiliated milblogger criticised the official response, warning that it eroded public trust and that the mobile restrictions do “not really help.”¹

Inferred, and labelled as inference by the people who drew it: that authorities appeared to prioritise State Duma election turnout over public safety.¹ ISW’s own words are that officials appeared to place a high turnout above civilian protection.¹ That is an assessment of motive from a pattern of behaviour, offered by an institution whose business is exactly that, and it is not the same thing as a document in which somebody wrote the order down. I have no such document and I am not going to write as if I did.

What would change my reading? An official explanation would — a technical failure in the alert system, a decision taken for some other stated reason, a denial with detail in it. None appears in the report I read. If one is produced I would want to weigh it.

The corroboration that impresses me most is not the analysis. It is the complaints. People in the Moscow region voiced anger over the silence, some of them on the governor’s own channel, and the channel disabled comments.¹ That is a reaction from inside the thing being described, by people who did not need an assessment to tell them something was missing. Then the place they said it was closed.

Here it is, stated once and without hedging, because the evidence above will carry it.

A warning is the one thing in this story that exists for no one but the person who receives it. Everything else that night served the state: the air defences, the restricted mobile service, the counted ballots. The siren and the text message serve the resident, and on this night, by ISW’s account, the resident is the one who did without.¹

Hold the comparison still, because it is the entire case. The warnings went out during significantly smaller attacks. They did not go out during this one.¹ Whatever reason a government might give for a silent alarm, it cannot be a reason that grows weaker as the attack grows larger. A system used for less and withheld for more has been repurposed, and it is no longer serving the people it addresses.

That is the moral, and it generalises further than Moscow. A state that manages the appearance of normality during an attack has made a choice about which of two things it needs more: citizens who are safe, or citizens who are calm on the day they are counted. Elections are supposed to be the moment a government is most accountable to the people. This is the inversion — the vote used as a reason to tell them less.

And I would ask any reader inclined to file this under things Russians do to Russians to hold the thought a moment longer. The mechanism has nothing Russian in it. The mechanism is: the government controls the alarm, the government is being judged that day, and nobody is in a position to make it ring. We have written about official guidance in Brussels and Helsinki telling households to hold 72 hours of supplies, and about a Czech spy chief who says a limited Russian incursion into NATO territory could come in months, not years while Latvia’s service sees no indication of an imminent attack.² Preparedness guidance assumes a warning will reach you. It is worth knowing that warnings are discretionary, and that the discretion belongs to whoever has a bad week to get through.

The strike was not a mystery raid. Ukraine’s General Staff reported that the barrage hit the Gazpromneft-Moscow Oil Refinery in Moscow City, about 450 kilometres from the Ukrainian border, starting a large fire and damaging its primary oil-processing unit.¹ The refinery is among Russia’s largest, with an annual capacity of about 12 million tons, and the Ukrainian manufacturer FirePoint says its FP-1 drones forced the plant to pause refining and lifted the share of offline Russian refining capacity from 10 to 14 per cent.¹ That is a company’s account of its own product’s effect, and I record it as one. It belongs to the campaign this desk has been tracking through the Kapotnya refinery fire and Moscow’s claim that it downed 450 drones on election day.²

ISW assessed that Ukrainian forces did not deliberately target polling stations or other electoral infrastructure.¹ I note that because it removes the one excuse that would have made the silence defensible: nobody can argue the sirens were withheld to avoid panicking voters away from a threat aimed at voters.

All of it sits inside a diplomatic frame we have covered before. The strikes continue in the absence of Russian agreement to the energy ceasefire Trump proposed on 14 September, and UNITED24 reports that Zelensky has repeatedly reaffirmed Ukraine’s willingness to accept a mutual moratorium on energy strikes while Putin has rejected each proposal.¹ That is the Ukrainian outlet’s characterisation of the Russian position, and readers should weigh it as such. We have covered the offer Zelensky made to stop separately.²

So, the ledger. A research institute assesses that a government withheld an alarm it has sounded for less, on the night it was being voted on.¹ Its own residents noticed and complained, and the channel where they complained closed its comments.¹ A Kremlin-affiliated milblogger said the response eroded public trust.¹ I cannot show you the order and I have not pretended to. What I can say is what the report shows: the warning existed, had been sent during smaller attacks, and was not sent during this one.¹ When residents said so on the governor’s channel, the comments were disabled.¹ Of those two facts the second is the one I would put to any official willing to answer, because a government confident in its reason for the silence has no cause to switch off the people asking about it. That last sentence is my judgment. The silence, the comparison and the closed comments are the report’s.¹