VIGIL CONSILIUM Search
Context. Not headlines.
← Analytical Notes

Analytical NoteSahel

Mali's regeneration problem: the positions that do not come back

The Malian army reinforces and holds its hubs. Since mid-2025 it has not been documented re-establishing a single outlying camp it lost — and at Dioura, on 10 September, no relief column was reported at all.

Published
2026-09-19
As of
2026-09-18
Frame
Strategic context
Confidence: Moderate

Key findings The note's claims, in the order the argument makes them.

  1. 01

    Since mid-2025, no overrun Malian position has been documented as re-established. The hubs are a different story — Sévaré, Gao, Konna and Somadougou were attacked and held — but an outlying camp that falls stays fallen.

  2. 02

    At Dioura the camp ended without anyone taking the ground. Local sources reported no relief column, the camp emptied of its weaponry, and the site still abandoned eight days after the assault of 10 September.

  3. 03

    JNIM does not occupy what it overruns, and has not needed to. It claimed total control of the barracks, withdrew, and left a vacancy that neither side has filled.

  4. 04

    The state's replacement channels are real and external — 24,000 recruits announced, Russian sealift through Conakry and Lomé, an eighth Akıncı in July. What has visibly contracted is air support: at least thirteen aircraft lost since 2023, and none over Dioura.

  5. 05

    The evidence cannot separate inability from consolidation. Reporting says no relief column was sent; no source says none could be, and Mali has published nothing about its losses or its posture.

What happened after the attack Seven cases from the last sixteen months, each read for the same three things: what the attack did, what the state did next, and where the position stands now. The verdicts are Vigil's; the basis column says what kind of evidence each row rests on.

DimensionThe attackWhat the state didThe position nowVigil's grade
Dioura, 23 May 2025JNIM overran the camp and the town; 41 soldiers reported killed; vehicles and weapons seized.No relief or recapture reported. The attackers left the same day.Re-garrisoned at some point before September 2026 — the only case here in which a position returned, and nothing documents how or when.ChangedReported — agency and compiled reporting
Boulikessi, 1 June 2025JNIM overran the Malian and Russian position in the Douentza cercle; about 100 soldiers reported killed, 22 captured; the barracks looted.A counterattack was mounted and failed on 5 June.Abandoned on 6 June 2025 and not re-established.ChangedReported — agency reporting
The far-northern posts, late April 2026The FLA took Kidal during the joint offensive; Malian and Africa Corps forces gave up Aguelhok, Tessalit, Tessit and Ber.Withdrawal, several of the posts reportedly under negotiated terms rather than after a fight.None re-established. The north of Mali is no longer a state sustainment area.ChangedCorroborated — already canonical in this module
Anéfis, 4–10 July 2026The FLA and JNIM took most of the town and besieged the base; a relief column from Gao was ambushed near Tabrichat and an Mi-24 was downed.A second column was sent and broke through, at a stated cost of about thirty soldiers.Retaken, the army announced on 10 July. The base was besieged, not overrun — which is why this is the counter-case and not a counter-example.ContinuityCorroborated
Konna and Somadougou, 4 July 2026Attacked in the same wave that struck Gao, Sévaré, Aguelhok and a prison near Bamako.Defended in place. The army said both were held with the help of Africa Corps — a rare public acknowledgement of the partnership by the Malian army itself.Held.ContinuityReported — the army's own statement
San, 6–9 August 2026JNIM attacked the garrison in the Ségou region; ten soldiers and forty attackers reported killed, and the organisation later claimed the barracks.The army said the attack was repelled.Unresolved. The two accounts have never been reconciled and no independent verification has been published.UnknownContested — two irreconcilable accounts
Dioura, 10 September 2026JNIM overran the camp and claimed total control of the barracks; tolls given to AFP rose over eight days from more than fifty to more than a hundred killed, with five Africa Corps personnel among them and between sixty and ninety captured.Air-ground operations were announced against the attackers. Local sources said no reinforcements were sent to the site and that the attackers returned the following day.Emptied of its weaponry and still abandoned eight days later, with no administrative presence documented by anyone.ChangedReported — agency reporting from security, military and local sources; no official figure

ChangedPartly changedContinuityUnknownSeven cases is a pattern, not a dataset. There is no published list of Malian positions, so this table cannot claim to be complete, and a camp quietly reoccupied would not necessarily have been reported. The May 2025 Dioura row is included precisely because it cuts against the reading: the camp did come back, and nothing explains how.

Positions attacked, held and given up One question only: where the state defended a position successfully, and where a position ended. A marker is a place whose status is established by published reporting — never a garrison, a strength, a route or a current disposition, none of which this module publishes. Nothing is drawn between the markers, because nothing in the evidence supports a line.Open the interactive map

Positions attacked, held and given upBamakoSévaréKonnaDiouraGaoAnéfisKidalAguelhokTessalit
  • Confirmed — official statement
  • Reported
  • Held, then left
  • As of 2026-09-18
  • BamakoConfirmedThe capital, supplied by escorted fuel convoy in the blockade's twelfth month.
  • SévaréConfirmedThe central hub. Attacked in April and July 2026 and held; everything in the centre is reached from here.
  • KonnaConfirmedAttack of 4 July 2026 announced as repelled, with Africa Corps support acknowledged by the army.
  • DiouraLeftOverrun 10 September 2026, reported emptied of its weaponry, still abandoned eight days later.
  • GaoConfirmedThe northern hub, attacked on 4 July 2026 and held; the origin of both Anéfis relief columns.
  • AnéfisReportedBase besieged in July 2026, first relief column ambushed, town announced retaken on 10 July.
  • KidalLeftTaken by the FLA in the April 2026 offensive; not returned to.
  • AguelhokLeftGiven up in the April 2026 withdrawals, reportedly under negotiated terms.
  • TessalitLeftGiven up in the April 2026 withdrawals.

Basis: Sources: Established international media reporting · France 24 / RFI · ACLED conflict data & analysis · Al Jazeera English

Objects in this note The canonical records the argument rests on. Each opens the record, with its own sources, dates and grade.View all in Explore

Key judgment

A working question about JNIM's pressure on Malian sustainment survives this research only in a narrowed form. The evidence does not support a general claim that the state can no longer supply what it holds: the hubs are reinforced and held, the capital is supplied by escorted convoy in the blockade's twelfth month, and a besieged garrison at Anéfis was relieved in July. What the evidence does support is narrower and sharper. Since mid-2025 no overrun Malian position has been documented as re-established, and the camp at Dioura — attacked for at least the fourth time since 2019 — was reported emptied of its weaponry, left without a relief column, and still abandoned eight days later. That is a position ending without anyone holding the ground it stood on. It does not mean the state is collapsing, that JNIM designed a campaign around regeneration, or that the pattern is a limit rather than a consolidation taken deliberately. Nobody has published enough for that last question to be answered.

The finding

The question this Note set out to test was whether JNIM is generating strategic pressure by degrading the Malian government’s ability to sustain and regenerate fixed positions, rather than by holding the ground it attacks. The research does not support that claim in its general form, and it supports one half of it strongly.

The half that does not survive is sustainment. Mali’s hubs are supplied and defended. Sévaré and Gao were attacked in April and again on 4 July 2026 and held. Konna and Somadougou were attacked on the same day and held, and the Malian army said publicly that it held them with the help of Africa Corps. A base besieged at Anéfis in July was relieved from Gao after the first column was ambushed and an Mi-24 was lost, and the town was announced retaken on 10 July at a stated cost of about thirty soldiers. The capital has been supplied by escorted fuel convoy for twelve months against a declared blockade. None of that is the behaviour of a force that can no longer move.

The half that survives is regeneration, and it is narrower and more specific than the question assumed. Since the middle of 2025 no overrun Malian position has been documented as re-established. Boulikessi was overrun on 1 June 2025, a counterattack failed on the 5th, and the base was abandoned on the 6th. The far-northern posts given up in late April 2026 — Aguelhok, Tessalit, Tessit, Ber — have not been returned to, and neither has Kidal. And on 10 September 2026 the camp at Dioura, in the Ténenkou cercle of the Mopti region, was overrun by JNIM, which claimed total control of the barracks on its own channels. Local sources told AFP that the attackers came back the following day, that the army sent no reinforcements, that the camp was emptied of all its weaponry, and that it was still abandoned eight days later.

That last case is the clearest instance in this module of a mechanism the conflict’s vocabulary handles badly. Nobody took Dioura. JNIM did not stay, the state did not return, and the position simply stopped existing. The map changed without a battle for the ground having been fought or won by anyone.

What a position needs in order to come back

A garrison is not a place; it is a set of dependencies, and an attack that breaks enough of them at once removes the position whether or not the attacker remains. Seven things have to be true for an overrun camp to be re-established, and the September case failed at least four of them at the same time.

Personnel. The dead and captured have to be replaced with people who can hold the post. Mali has a stated programme for this: the 2025–2026 government action plan announced the recruitment of 15,000 military and 9,000 security personnel, together with camp construction, equipment purchases and road surveillance. What the plan produces is not audited publicly. What is known about the specific case is that a security source in Mopti told AFP that many of the soldiers at Dioura were on their first field deployment.

Equipment. A camp that is looted rather than destroyed transfers capability instead of merely losing it. An open-source compilation of visually confirmed Malian losses from the April 2026 fighting counted 137 items, of which 99 were assessed as captured and 38 as destroyed — a floor rather than a total, and an open-source one, but the ratio is the point. At Dioura the camp was reported emptied of its weaponry. Replacement runs increasingly through a sea channel: a Russian cargo ship unloaded military vehicles at Conakry in March 2026 for onward movement to Mali, and a second delivered 29 armoured vehicles through Lomé in July.

Fuel. Everything above moves on imported diesel, and Mali imports all of it. The corridors from Senegal and Côte d’Ivoire carry the overwhelming majority; JNIM declared a blockade of them in September 2025, more than three hundred tankers have been reported destroyed, the government imposed nationwide rationing in January 2026, and a full blockade of the capital was reimposed on 28 April. The corridor is degraded and taxed rather than closed — over a thousand tankers are reported to have reached Bamako under escort since late October 2025 — but the fuel that would move a relief column is the same fuel that arrives under escort, and it is not abundant.

Movement. The delta is the constraint the map does not show. Dioura lies about 130 kilometres west-north-west of Sévaré in a straight line, and considerably further by road, across an inland floodplain that fills roughly from September to December. This module’s terrain record has described that seasonal effect since it was written. What no source establishes is whether it mattered on 10 September, and this Note does not assert that it did.

Reinforcement. A column has to be ordered, assembled and sent. At Anéfis in July one was, twice. At Dioura in September, on local accounts, none was.

Command. The portfolio that owns the army’s reconstitution has been held by the head of state himself since 4 May 2026, when Assimi Goïta took the defence ministry by decree nine days after Sadio Camara was killed in the April offensive; the army chief of staff was made minister delegate at the same time. There is no separate minister to carry a loss, and no official toll has been published for the heaviest one this module records.

Air support. This is the dependency that has visibly contracted. Open reporting counts at least thirteen modern manned and unmanned aircraft lost by Mali since 2023, including all four of its Mi-24P attack helicopters, the last of them downed during the July fighting at Anéfis. Air assets are what let a state watch a road, escort a column and support a besieged camp — and the soldiers at Dioura, according to the same Mopti source, had none.

Why a raid that ends in withdrawal still removes a position

The standard test of an insurgency’s progress is whether it holds what it takes. On that test JNIM has made almost no progress at all: this module’s Watch record has found no incursion followed by administration in any review, and Dioura is now the strongest case of that, because there the ground was vacant and still nobody occupied it.

But the test measures the wrong thing when the objective is subtraction rather than acquisition. A group that destroys a garrison’s manpower, removes its weapons, kills or captures the people who would have formed the core of its replacement and then leaves has imposed the cost of the position on the state permanently, while carrying none of the cost of holding it. It does not have to govern, feed, defend or supply anything. If the state does not come back, the outcome is identical to occupation, and it is cheaper.

Analysts have described the operational grammar of this without naming the effect. The Sahel researcher Daniele Garofalo characterised JNIM’s approach in July 2026 as fighting “through mobility, repetition, and selective escalation” in a way that exposes a structural mismatch for state forces, which are pushed into reactive postures before ground or air reinforcements can engage. The Institute for Security Studies described the same organisation’s corridor campaign in June 2026 as a strategy of asphyxiation aimed at economic cost rather than territory.

Repetition is the part that matters here. Dioura’s camp was raided in March 2019, overrun by northern movements in September 2023, taken and looted by JNIM in May 2025, and taken again in September 2026 — four attacks on one position in seven years, each ending in withdrawal. A position attacked that often is a position whose replacement cost the state pays over and over, and the 2026 instance is the first in the sequence where it appears not to have paid it.

None of which establishes intent. JNIM has published no statement describing a campaign against Malian force regeneration. Its stated demands concern religious rule and the departure of foreign forces, and its declared blockade was framed around fuel taxation and sharia. The pattern here is Vigil’s reading of what the organisation does, not a claim about what it has decided.

The geography

Central Mali is held from one place. Sévaré, beside Mopti, carries the central military region’s seat, the airfield from which the region’s air operations run, and the point where the road network of an entire region narrows. Konna, about 60 kilometres north-east of Mopti on the eastern edge of the floodplain, is where the road to Gao leaves the delta — and is where a jihadist column was stopped in January 2013, the event that brought France into the war within hours. Everything that reaches a camp in the delta starts at or passes through this pair of points.

That concentration is why the pattern reads the way it does. The hubs are defensible because they are where the force is, and they have been defended. The camps are hard to relieve because they are at the end of a delta road from a single origin, and since mid-2025 they have not been relieved. Reading the centre as a contest over towns misses this entirely; reading it as a contest over what a single hub can reach, and what it chooses to reach for, explains both halves of the record.

The same concentration explains the state’s substitute. As ground positions contracted through 2026, strike became the instrument by which the government asserted reach into places it no longer occupied. On 16 September 2026 Bellingcat published a geolocated casework of Malian and Africa Corps drone strikes in the north, describing a recurring dynamic its investigators call “guilt by location” — people treated as targets on the basis of where they were or what they were moving, fuel and mining among the activities named, rather than on established combatant status. That is the cost side of the substitution, and it belongs in this Note rather than in a separate file, because a drone can strike a place the state no longer holds and cannot reopen a camp there.

What the partner does and does not change

Africa Corps is present at the point of failure as well as at the hubs. Five of its personnel were reported among the dead at Dioura, which is a small fact with a structural implication: the Russian presence in the centre is not only aviation and escort at the regional capitals. It also changes some outcomes — the Malian army’s own statement on 4 July credited Africa Corps with holding Konna and Somadougou, and a Russian column was part of the Anéfis relief.

What it does not change is the mechanism this Note describes. Africa Corps is sustained on its own external chain and supplies presence, fires and escort. It does not generate Malian soldiers, and it did not put a Malian camp back. Moscow said nothing about the five deaths at Dioura, as it has said nothing about any Russian loss in Mali this year.

What this does not establish

That the army could not have relieved Dioura. Reporting says no reinforcements were sent. No source says none could be sent, and the difference is the whole argument. A force with finite manpower that abandons a camp 130 kilometres into a flooding delta and concentrates on a hub it can hold is making a defensible choice, not failing. Both readings fit every fact in this Note, and the evidence that would separate them — an order, a stated posture, or a return — is not public. This is the strongest objection to the reading above, and it is not answered here.

That the pattern is complete. Mali publishes no list of its positions. A camp quietly reoccupied would not necessarily be reported, and the May 2025 Dioura case shows this directly: the camp was overrun and evidently re-garrisoned at some point before September 2026, and nothing in the open record says when or how.

That the casualty figures are settled. The Dioura tolls given to AFP rose from more than fifty to more than a hundred over eight days, from different sources describing different things — an officer at the scene, a security source in Mopti, a local official, a military police report. JNIM claimed 150 killed and 93 captured. None of these is an official figure, because Mali has published none.

That this is the war’s direction. No Malian regional capital has fallen. The front of pressure has not moved south since July. A pattern fifteen months long across a handful of camps is a finding about those camps and about a mechanism; it is not a trajectory, and this Note does not publish one.

That JNIM is winning by this. The organisation has demonstrated it can remove positions. It has not demonstrated that it can replace them with anything, and the vacancy at Dioura — unoccupied by either side eight days on — is as much an observation about the limits of its project as about the limits of the state’s.

Source basis and method

This Note rests on four kinds of evidence, kept apart throughout. Corroborated: the April 2026 withdrawals and the July relief at Anéfis, which are already canonical in this module and carried by several outlets; the 4 July attacks and the Malian army’s own statement about Konna and Somadougou. Reported: every casualty figure, every account of what happened at Dioura after the attack, the Boulikessi sequence, the equipment-loss compilation and the Russian sealift deliveries — each attributed, most of them to one agency’s security, military and local sources. Claimed: JNIM’s account of Dioura, carried as a claim. Rejected: two things worth naming. Secondary summaries routinely date the heaviest attacks on escorted tanker convoys in western Mali to September 2026; they occurred in September 2025, and no distinguishable September 2026 tanker attack was found. And no location record was created for Boulikessi, because no source consulted gives coordinates for it that Vigil would be willing to draw.

The pattern across the cases is Vigil’s reading and is graded moderate for that reason. The individual cases are not in dispute; what they add up to is an argument, and the section above says what would break it.

Conclusion

What this means

  • The measure of the Malian state's military position in the centre is no longer which camps are garrisoned but which garrisons return after they are lost. On the documented record since mid-2025, the answer is none, and the interval is now fifteen months long.
  • An armed group that can make a position unreplaceable gets the same map as one that captures ground, at a fraction of the cost and with none of the obligations. JNIM has behaved throughout as though it understands this, whatever its stated aims are.
  • The state's replacement of men and machines is increasingly external — announced recruitment it does not audit, and armour that arrives by Russian ship through West African ports. What no partner supplies is the decision to go back to a place.

What this does not mean

  • It does not mean the army cannot supply what it holds. The hubs were attacked and held, a besieged base at Anéfis was relieved, and the capital has been supplied by escorted convoy for a year.
  • It does not mean JNIM has a regeneration campaign. Nothing it has published says so; its stated demands are religious and political, and the pattern is Vigil's reading of its behaviour, not its plan.
  • It does not mean the state is collapsing. No regional capital has been lost, and the front of the war has not moved south since July.
  • It does not mean Dioura proves inability. Reporting says no relief column was sent. No source says none could be, and a force consolidating deliberately would produce exactly the same record.

What to watchWatch on Sahel

  • A return to Dioura, or to any comparable vacated site — troops in place rather than a clearing operation announced. It would be the first since mid-2025 and would change this assessment more than any other single fact.
  • Whether a Malian statement of posture is ever published — an acknowledgement that outlying camps are being consolidated by intent, which would settle a question the evidence currently cannot.
  • The interval between outlying positions overrun and not returned to. The gap is the measure; it does not need a new incident to be informative.
  • Whether a hub ever needs relief from outside its own region. That would move the question from what the state does not rebuild to what it cannot hold.
  • Whether the vacated ground is administered by anyone. It was not at Dioura eight days on, and an insurgency that starts governing what it empties is a different problem from one that raids.

Explore Sahel

This note explains one development. The standing coverage holds the system it sits inside — the map, the actors, the forces, the history and the economic and humanitarian conditions — each with its own sources, dates and confidence.

Open SahelUnderstand this conflict in Explore

Sources and confidence

Confidence: Moderate The individual cases are corroborated: the same agency reporting is carried by several outlets, the attack dates are consistent across accounts, and the April 2026 withdrawals and the July relief at Anéfis are already canonical in this module. The pattern across the cases is Vigil's reading and is graded moderate for that reason. Every casualty figure here is attributed reporting from security, military or local sources, usually to one agency, and the Dioura range — fifty to more than a hundred — is the width of the uncertainty rather than a sequence of corrections. Mali has published no toll, no order of battle and no statement of posture, so the claim that no position has been re-established rests on the absence of reporting about small garrisons, which is a thin kind of evidence and is graded no higher than the visibility of the thing.

Sources: Established international media reporting · France 24 / RFI · ACLED conflict data & analysis · Al Jazeera English · International Crisis Group · Institute for Security Studies (ISS Africa) · Africa Defense Forum (ADF) · Africa Center for Strategic Studies · Military Africa · Bellingcat open-source investigations · Armed-group statements (JNIM, FLA, IS-Sahel and others) · Official government statements · UN OCHA humanitarian reporting (Mali / Sahel)

How Vigil grades and dates claims is set out in the methodology. Spotted an error? See the corrections log.

Vigil Dispatch

Follow the analysis

Get Vigil’s Weekly Briefs and new Analytical Notes by email.

One confirmation email, then Weekly Briefs and new Analytical Notes. Unsubscribe at any time.