What a quarter-million deserters do, and do not, tell us about how this war ends

Ukrainian AWOL and desertion cases together approached 290,000 by September 2025, the last figures published before Kyiv classified the data.

By the Eastern Europe Desk · May 2026 · 13 min read

Ukrainian AWOL and desertion cases together approached 290,000 by September 2025, the last figures published before Kyiv classified the data. Russian AWOL prosecutions are on pace for 70,000 in 2025. The comparative-conflict literature treats numbers on this scale as a symptom of military strain, but cautions that desertion is rarely a clean predictor of when a war ends.

The Ukrainian Prosecutor General's Office published quarterly statistics on criminal cases involving the offences of *unauthorised absence from duty* (Article 407 of Ukraine's Criminal Code) and *desertion* (Article 408). Both offences have specific legal definitions and prosecutorial thresholds, and both have been the subject of significant criminal-justice activity since the February 2022 invasion. The cumulative figures through September 2025, drawn from the Prosecutor General's published data and from monitoring by independent Ukrainian outlets, show 235,646 cases of unauthorised absence from duty and 53,954 cases of formal desertion since February 2022. The combined total is approximately 290,000 individual military personnel against whom one of the two charges has been formally registered.[1][2]

The Russian comparator is harder to assemble because the Russian state has not published equivalent figures, but the Mediazona project, which catalogues Russian court records exhaustively from publicly available court databases, has identified 10,308 cases of *refusal to serve* (Article 332.2.1) and 5,204 cases of *AWOL* (Article 337) prosecuted in Russian military courts in 2024 alone, with the full 2025 totals expected to reach approximately 70,000 if the first-half-of-2025 pace continues. Cumulatively across the war, Mediazona's catalogue identifies approximately 16,000 criminal cases through 2024, with 14,182 of those involving AWOL. The Russian figures are an undercount; the Ukrainian figures, while they were published, were closer to comprehensive, but the case-management threshold for what counts as a recorded AWOL case differs substantially between the two systems, making a one-to-one comparison unreliable.[3][5]

The aggregate magnitudes nonetheless support a clear quantitative claim: the combined Russian and Ukrainian desertion-and-AWOL cohort across the post-2022 war is comfortably in excess of a quarter-million people. The available evidence does not establish whether the true cumulative figure is closer to 300,000 or 500,000, because both states are operating under conditions in which the criminal-justice response to military personnel absence is selective rather than comprehensive. What is established is that the combined total runs to the hundreds of thousands, and that the rate of new cases accelerated across 2024-25 rather than stabilising.

The comparative-conflict literature gives this figure some analytical meaning, though less than is sometimes claimed for it. The political-science literature on military cohesion is clear that no published study establishes a single numeric desertion rate above which a war reliably ends; desertion is treated as a *symptom* of military and political strain rather than a tripwire. Jasen Castillo's *Endurance and War* (Stanford University Press, 2014) is the standard recent treatment of why some armies hold together in protracted war and others come apart. His argument is structural: military cohesion is determined by two variables, the degree of *regime control* a state exercises over its society, and the degree of *autonomy* its armed forces have to build internal bonds and focus on warfighting. Castillo's typology runs from "messianic" and "authoritarian" armies, which sustain cohesion under extreme strain, to "apathetic" armies, which do not. His finding that non-democratic states often field the more durable forces in long wars, and that democracies face cohesion disadvantages in protracted conflict, is the part of his work most relevant here; his book contains no rate-based threshold.[6]

The historiography of mass desertion is, if anything, a caution against reading the case counts as a forecast.

The Ukrainian case

The Ukrainian Armed Forces have approximately 800,000 total personnel as of late 2025, including approximately 700,000 in active combat units and the remainder in training, support, and territorial defence roles. The 235,646 AWOL cases plus 53,954 desertion cases represent approximately 290,000 individual personnel, roughly 35 percent of the total active force, counting cumulatively since the war began.[1]

The figure is misleading taken at face value. Not all of the 290,000 are still missing; many have returned through the November 2024 amnesty programme (more than 29,000 returnees between November 2024 and August 2025), through formal prosecution and reintegration into reduced-status units, or through informal return-to-unit processes that did not produce a criminal-justice resolution. The cumulative figure also includes second-time and third-time absentees, who are counted as separate cases. The Frontelligence Insight analysis judges that the case totals roughly double the number of distinct people involved, and estimates the true net loss to the force at on the order of 150,000 personnel.[2] Set against an active force of roughly 800,000, that is in the high-teens as a share of strength, a calculation we make here from those two figures, not one Frontelligence states. It is a large number by any measure, but it is an order-of-magnitude estimate built on contested case-counting, not a precise rate.

The trajectory is the concerning element. The pace of new cases has accelerated sharply: through September 2024 the cumulative count of AWOL and desertion cases stood at roughly 89,000 (59,606 AWOL plus 29,521 desertion), but roughly 200,000 *new* cases were added in the single twelve-month window between September 2024 and September 2025, more new cases in one year than in the war's first two-and-a-half years combined.[2] The acceleration reflects the cumulative effect of more than three years of intensive combat, the deterioration of unit cohesion in long-engaged formations, the introduction of expanded conscription that has brought less-motivated personnel into the force, and the general exhaustion of a war whose end is not in sight.

The public record now ends there. October 2025 set a single-month record of 21,602 new AWOL and desertion cases, and shortly afterwards the Prosecutor General's Office stopped publishing the monthly series, confirming in December 2025 that the data had been reclassified as restricted information on the grounds that open statistics gave Russia insight into discipline, staffing levels, and the effectiveness of mobilisation.[13][14] Since the blackout, the official sense of scale has come in ministerial statements rather than data: Defence Minister Mykhailo Fedorov said in January 2026 that roughly 200,000 soldiers were absent without leave.[13]

The Zelensky government's response, the November 2024 amnesty programme, was an explicit acknowledgement of the magnitude. The amnesty's terms (first-time deserters can return to their units without criminal prosecution) produced more than 29,000 returnees by August 2025, a useful but modest fraction of the cumulative cohort.[13] The amnesty has not been extended to second-time deserters, formal desertion (as opposed to AWOL) cases, or to personnel who left before specific operational dates; the residual cohort that remains outside military authority is substantial.

The Russian case

The Russian Armed Forces' total deployed strength has been more contested. The most reliable estimates put Russian deployed strength in the Ukrainian theatre at approximately 600,000-700,000 across 2024-25, with broader Russian military personnel under arms substantially larger. The cumulative AWOL and desertion case count of approximately 16,000 through 2024 represents a relatively modest 2-3 percent of deployed strength on the catalogued cases alone.[3][8]

But Mediazona's case count is, again, an undercount. The Russian criminal-justice system has applied AWOL and desertion charges selectively, typically against personnel who have left and refused to return for extended periods, rather than against personnel whose absences are administratively resolved through pressure or family-intervention. Independent analysis of the Russian case suggests the true desertion-AWOL cohort is materially larger than the catalogued cases, on the order of tens of thousands of personnel and rising through 2025.[7] Measured against a deployed strength of 600,000-700,000, even the higher independent estimates put the Russian rate well below the Ukrainian share, a gap consistent with the structural argument that a tightly controlled, coercive force suppresses the *visible* exit rate without necessarily resolving the underlying strain. Precise percentage comparisons between the two armies are unreliable: the two states count cases on different thresholds, and the conversions here are our own, built on contested estimates.

The Russian state's response has differed structurally from the Ukrainian state's. Russian military authorities have relied on a combination of severe criminal-justice penalties (sentences of 5-10 years for AWOL, life sentences for desertion in wartime), public-information campaigns identifying deserters by name, and informal violence, including documented cases of deserters being executed in penal battalions. The Russian state has not offered amnesty terms equivalent to the Ukrainian government's November 2024 programme. The combination of severe formal penalties and continued unit pressure has compressed the visible desertion rate without resolving the underlying cohort, and the Mediazona case-count growth (10,308 cases in 2024, vs 5,517 in 2023) reflects the increasing willingness of Russian commanders to escalate cases through the criminal-justice system rather than resolve them informally.[3]

The historical record

The temptation with numbers this large is to read them as a countdown. The historical scholarship does not support that reading. What it supports is more modest and more useful: that sustained, accelerating desertion is a reliable *symptom* of an army under deep strain, and that in the cases where it has run to genuinely mass scale it has tended to be a constituent part of a broader collapse rather than its sole cause.

The clearest case is the one the present numbers most often evoke: the Russian Imperial Army in 1917. In the standard scholarly treatment, Eric Lohr and Joshua Sanborn estimate that roughly two million soldiers deserted between March and October 1917, a "self-demobilization" that they argue directly precipitated the collapse of the state's monopoly on force and Russia's exit from the First World War.[9] But the lesson historians draw from that episode is precisely *not* that a desertion rate crossed a tripwire. As the 1914-1918-Online encyclopedia's survey of soldiers' attitudes makes plain, desertion across all the major belligerents was a symptom of war-weariness that rose alongside, not ahead of, political breakdown, and scholars caution explicitly against treating desertion rates alone as an explanation for why an army stopped fighting. The German army's 1918 dissolution, on this account, came more through officer-ordered surrenders than through individual flight, even though some historians have read its late-war desertions as a "covert military strike."[11]

That caution cuts directly against the cleaner story. There is no published study establishing a percentage of force above which a war reliably ends, and the analyst who claims one is over-reading the evidence. What the literature does offer is a framework for thinking about *why* one army's strain converts into political pressure faster than another's. Dan Reiter's *How Wars End* argues that states move toward settlement chiefly as new information shifts their reading of the balance of power and of the enemy's resolve, and as commitment problems are resolved or worsen.[12] Rising desertion feeds that information channel as a visible signal of eroding resolve and reserves, but it is one input among several, mediated by what the political leadership does with it.

This is where Castillo's structural argument earns its place. His claim is that regime control and military autonomy, not headline desertion counts, govern how much strain a force absorbs before it breaks, and that democracies are at a comparative disadvantage in protracted war precisely because they cannot enforce the absolute-loyalty norms that hold the most coercive armies together.[6] Read through that lens, the Ukrainian and Russian numbers tell a less symmetrical story than a side-by-side percentage would suggest: the same underlying strain may be showing up as recorded cases in the more open Ukrainian system (an openness the late-2025 classification decision has now curtailed) while being suppressed, through coercion, into a lower visible rate on the Russian side. Lawrence Freedman has made a related, narrower point in his *Comment is Freed* commentary on Ukrainian manpower, that the trajectory and the replacement rate matter more than any single cumulative figure.[10]

What the strain does and does not foreclose

What the evidence supports is narrower than a prediction, and it is worth stating plainly. Both armies are absorbing manpower strain on a scale that, historically, has accompanied the late phases of long wars, but neither set of figures forecloses a particular outcome or timeline. The cases that historians cite as cautionary examples cut both ways: external resupply has, at times, substituted for manpower losses long enough to keep a strained force in the field (as with U.S.-supplied South Vietnamese formations), and authoritarian coercion has compressed the political effect of even very high desertion for extended periods (the Soviet experience in Afghanistan, and indeed Russia's own war effort to date). Mass desertion did help bring down the Russian army in 1917, but only once political authority had already fractured, which is not the present situation on either side.

Both Ukraine and Russia have mitigations that have, so far, held. Ukraine's external support (EU financial assistance, residual U.S. and European weapons supply, and broader Western political backing) has sustained the war effort despite the manpower pressure. Russia's coercive control over its own force has held the visible desertion rate down even as the catalogued case count rises. Neither mitigation is guaranteed to be permanent, and the open analytic question is how long each can keep absorbing strain that is still accelerating.

**Forecast.** *On the trajectories visible at the end of 2025, the most likely path is continued attrition rather than a desertion-driven collapse on either side over the next twelve to eighteen months: both states retain the means (external resupply for Ukraine, coercion for Russia) to keep replacing losses faster than desertion erodes the force. A sharper inflection becomes more plausible if either mitigation fails: a serious interruption to Western financing or matériel for Ukraine, or a loosening of regime control or a mobilisation shock for Russia. We assign no date to that inflection; the historical record gives no basis for one. This paragraph is a forecast, not a finding.*

The quarter-million-plus desertion-and-AWOL cohort is not, by itself, a prediction of war termination, and it should not be sold as one. It is a quantitative marker that both forces are under sustained manpower strain, a symptom the comparative-conflict literature treats seriously and reads cautiously. The political-leadership response to that strain, the willingness to negotiate, to absorb further casualties, to bear the continuing criminal-justice burden, will matter far more than the case counts themselves. The figures are real, large, and, until the public record was closed, accelerating. What they justify is closer attention to the political end-game question, not a countdown.

Open in the interactive site →