Sustainable War, Tested on Humans — Haiti

Abstract

Between March 2025 and the first quarter of 2026, the international response to Haiti’s gang crisis crossed a threshold that has gone largely unexamined: the forces deployed to protect the population became the leading cause of casualties among it. United Nations reporting for January through March 2026 attributes more than two-thirds of all killed and injured to security-force operations against gangs, a share that inverts the conventional picture of a population menaced chiefly by criminals.

This paper argues that Haiti has become the clearest contemporary instance of what the Ghost in the Shell: SAC_2045 writers called “sustainable war”: armed conflict reorganized as a self-financing enterprise, decoupled from any political settlement, and increasingly delegated to private military firms operating outside the accountability structures that constrain state forces. It traces the sequence from the failed Multinational Security Support mission, through the contracted drone campaign run by a Haitian government “Task Force,” to the newly authorized Gang Suppression Force, and shows how the incentive architecture at each stage rewards perpetuation over resolution.

The thesis is not that anyone set out to test weapons on civilians. It is that a system has assembled itself in which doing so carries no cost, whether legal, financial, or political, and in which the absence of collateral-damage discipline is a feature of the delivery model rather than an accident of it.

01 — The theater and the thesis

On 16 June 2026, United Nations Secretary-General António Guterres made a one-day visit to Port-au-Prince. According to reporting from the scene, as he met privately with Prime Minister Alix Didier Fils-Aimé, a displaced man outside the building began striking its metal siding and shouting that he wanted to go home, his voice rising until security moved the Secretary-General out.¹ The image is almost too neat: the figure of global stewardship ushered away from the sound of the governed.

The visit was occasioned less by the years of accumulated suffering than by a single recent event. A week earlier, James Boyard, the cabinet director of the Defense Ministry, had been abducted in one of the few relatively secure districts of the capital.² When the violence reached the senior administrative class, the calendar moved. That sequencing is itself a small piece of evidence for the argument that follows.

This paper is about a larger and less visible inversion. The dominant narrative of Haiti holds that an overwhelmed population is preyed upon by gangs while the world debates how to help. By early 2026 the data no longer supported that framing. According to the UN Integrated Office in Haiti (BINUH), in the first quarter of 2026 at least 1,642 people were killed and 745 injured, and more than 69 percent of those casualties resulted from operations carried out by security forces against gangs, with gang members responsible for roughly 27 percent and self-defense groups for the remaining 4 percent.³ The protectors had become the primary source of harm to the protected.

I will argue that this is not a story of a well-intentioned mission gone wrong. It is the predictable output of a particular model of conflict, one in which war is organized to sustain itself, in which lethal capability is leased rather than commanded, and in which the population of a collapsed state functions, in effect, as the proving ground on which the model is validated and refined.

02 — What “sustainable war” means

The phrase comes from Ghost in the Shell: SAC_2045, the 2020 series in which, following a global economic collapse, the major powers stabilize their economies by maintaining a permanent, managed, low-intensity state of conflict. It is war waged not toward victory but as an ongoing industrial activity, executed substantially by private military contractors and optimized to never quite end. The dystopian core of the idea is the alignment of incentives: once war becomes the economy rather than an interruption of it, the actors running it have no reason to conclude it.

Stripped of the science fiction, the concept names a real and documented trajectory in the study of armed force. Scholars of private military companies have charted the re-emergence of war as a commercial market, among them P. W. Singer in Corporate Warriors and Sean McFate in The Modern Mercenary and The New Rules of War. In their account, states purchase combat outcomes from firms whose business interest lies in the continuation, not the resolution, of the conditions that generated the contract. The crucial structural feature is the same one the fiction dramatizes: the entity delivering force is paid in a way that does not reward making itself unnecessary.

A salaried army wins and goes home. A contracted one is retained. The difference is not moral character; it is the shape of the invoice.

Three properties define a sustainable-war configuration, and Haiti exhibits all three. First, decoupling: the use of force is severed from any plausible path to political settlement, so that operations continue while the underlying drivers of conflict remain untouched. Second, delegation: lethal capability is exercised by actors such as private firms, ad hoc “task forces,” and non-UN coalitions, all positioned outside the legal and command structures that would otherwise impose collateral-damage estimation, proportionality review, and accountability. Third, self-financing: the arrangement is structured so that the conflict, or the institutions of the state it captures, pays for the force prosecuting it. The sections that follow examine each property as it appears in the Haitian case.

03 — The laboratory: a state hollowed from within

Haiti has had no elected national officials since early 2023 and no president since the assassination of Jovenel Moïse at his residence in July 2021.⁴ Executive authority has passed through a Transitional Presidential Council whose mandate lapsed in February 2026 amid internal feuding, leaving the government of Prime Minister Fils-Aimé to operate under contested legitimacy while elections are promised for later in 2026.⁵ Into that vacuum the armed groups expanded until, by most estimates, the coalition known as Viv Ansanm controlled roughly 85 to 90 percent of the capital.

The most consequential point for this analysis, and the one most often omitted from security-focused coverage, is what those groups actually are. The Global Initiative Against Transnational Organized Crime, in its March 2026 study Brokers and Patrons, argues that Haiti’s gangs are neither insurgents nor revolutionaries but actors embedded within the country’s circuits of political and economic power, sustained by illicit financial flows, arms and drug trafficking, and patronage networks that protect and instrumentalize them.⁶ They function, in the report’s framing, as brokers: nodes regulating access to territory, resources, and populations, deriving authority not only from violence but from their role as intermediaries between the state, its resources, and the communities they govern.⁷

The International Crisis Group documents how this relationship evolved. Outfits once tied to political parties grew militarily strong enough to shun their former patrons, demanding larger sums and pursuing their own objectives, with alliances becoming fluid and transactional.⁸ By late 2025 the leadership of Viv Ansanm was repositioning rhetorically as a protector of the poor against rapacious elites and even floating a political party, with the concrete aim of securing allies in the next administration and a general amnesty.⁹

This matters to the thesis because it identifies what a serious solution would have to reach. GI-TOC’s conclusion is blunt: arrests and targeted operations may weaken particular groups, but as long as the structures sustaining criminal governance remain intact, the system reconstitutes itself.¹⁰ A campaign that kills gang members while leaving the politico-criminal brokerage architecture untouched is, by construction, a campaign that cannot end the war. It can only run.

04 — The first intervention, and failure by design

The international response began conventionally. In October 2023, UN Security Council Resolution 2699 authorized a non-UN Multinational Security Support (MSS) mission, which Kenya agreed to lead, mandated to support the Haitian National Police.¹¹ Its concept of operations envisaged 2,500 personnel. It never came close.

Fewer than 1,000 personnel were ever deployed against the planned 2,500, and only a fraction of the required funding ever materialized. Yet the money that did flow was enormous. The United States alone channeled close to one billion dollars into the effort, more than US$970 million, and it is difficult to identify what that sum purchased.¹² There was scant public accounting of how the money was spent, few measurable objectives against which to judge it, and no durable outcome it can be said to have bought. The gangs that held the capital when the funds began arriving held it still when the mission ended. Nearly a billion dollars passed through the operation and effectively vanished, leaving behind neither secured territory, nor functioning institutions, nor any transparent record of where it went. This is the first and most under-examined feature of the Haitian intervention: an immense, opaque expenditure with no goals one can audit and no results one can point to.

The mission struggled with limited usable funding, legal constraints, and incomplete deployment throughout its existence, and armed groups continued to control much of the country.¹³ Kenya completed its withdrawal at the end of April 2026. Its officials framed the departure in the language of accomplishment, citing secured infrastructure and restored services, while outside analysts described a mission that had failed to curb the violence and that some judged bound to fail before it began.¹⁴

The relevant lesson is not that the Kenyans were incompetent. It is that the model was structurally starved: a police-support mission, legally confined to a supporting role, asked to contain a militarized criminal insurgency on a budget that arrived at perhaps a seventh of what was promised, while close to a billion dollars in associated spending dissolved without an accounting. When the legitimate, auditable tool is set up to fail at that scale, the illegitimate one acquires the appearance of necessity.

05 — The privatized turn: contracted air power over a city

While the MSS faltered, the Haitian government opened a second, parallel line of operation. From March 2025, a specialized “Task Force” established by Prime Minister Fils-Aimé began conducting drone strikes against gang targets. According to BINUH and other informed sources, it operated with the support of US private military contractors.¹⁵ Analysts at the International Crisis Group date the first use of so-called “kamikaze drones” to that month, as Haitian forces were rapidly losing ground and fears grew that the capital might fall entirely.¹⁶

The private military company most consistently named in connection with the Task Force is Vectus Global, led by Erik Prince, founder of the firm formerly known as Blackwater. Prince has stated publicly that Vectus signed a contract with Haiti’s transitional government and hired Salvadoran operators to support the Haitian police in using armed drones.¹⁷ Reporting on the agreement describes a structure that is, for the purposes of this paper, the most important single fact in the case.

Read against the framework in §2, the arrangement satisfies the self-financing property in almost literal form. The lucrative phase of the contract is not the year of fighting; it is the decade of sitting on the customs revenue of the state. A swift, decisive victory that handed a functioning government back to Haitians would not maximize the value of that position. An extended “stabilization” presence would. Whatever the intentions of any individual, the incentive vector points away from resolution. This is the mechanism the fiction imagined, rendered in the prose of a procurement agreement.

A portable model, not a one-off

The strongest evidence that Haiti functions as a proving ground rather than an isolated emergency is that the same operating concept appears elsewhere in the same firm’s portfolio. Reporting describes Vectus activity in the Democratic Republic of the Congo, where a February 2026 Reuters investigation reported that Prince had sent personnel and drones to assist the Congolese military alongside revenue-oversight work, and similar mercenary activity has been reported in Ecuador.²⁰ The tactics themselves are imported: the explosive-laden quadcopters used over Port-au-Prince are widely described as adapted from techniques of the war in Ukraine.²¹ A configuration in which frontline drone methods are imported, validated against a captive urban population under minimal oversight, and carried forward into the next contract is, in operational terms, a development pipeline. The product being refined is a model of cheap, deniable, contracted air war; the test range is a city.

These are human beings.

06 — Force without estimation: the human ledger

The defining technical characteristic of this campaign, and the one that most clearly marks it as something other than disciplined military operations, is the apparent absence of collateral-damage estimation. In professional targeting practice, a lethal strike against a target embedded among civilians is gated by a chain of deliberate steps: positive identification, a collateral-damage estimate calibrated to the weapon and the environment, a proportionality judgment, and post-strike assessment. That chain exists precisely to constrain the outcome now visible in Haiti. By every external indication, it was not applied.

Begin with scale. Human Rights Watch, drawing on multiple data sources, documented at least 1,243 people killed in 141 drone operations between 1 March 2025 and 21 January 2026, with 738 injured.²² UN reporting for the narrower window of 1 March to 20 September 2025 attributed at least 547 deaths to drone operations, of which 20 were members of the population, including 11 children.²³ By the first quarter of 2026, BINUH reported that of all victims of drone strikes in the period, 69 were members of the population, including five children.²⁴

Figure 1. Source of casualties, January–March 2026. Of at least 1,642 killed and 745 injured in the quarter, BINUH attributed more than 69% to security-force operations against gangs, 27% to gang members, and 4% to self-defense groups. The instrument deployed to protect the population had become its leading source of lethal harm. Source: BINUH, 8 May 2026.³²⁴

The qualitative record is consistent with an absence of estimation. HRW found no evidence that the people targeted posed an immediate threat to the lives of others, and documented repeated strikes on vehicles and on groups of people in densely populated areas.²⁵ A widely cited instance: in September 2025, four children playing in a street in Simon Pelé were killed by blasts from drones launched at a nearby party attended by gang members.²⁶ Earlier BINUH reporting noted that a substantial share of casualties during such operations were residents struck by stray fire at home or during daily activities, and that some officers had continued to summarily execute individuals suspected of gang links.²⁷

The decisive piece of evidence for intent, or its absence, is the posture of the authorities themselves. A Haitian government official, interviewed by the Washington Post, reportedly said that civilian casualties were to be expected.²⁸ That is not the language of a process that weighs proportionality and finds it satisfied; it is the language of a process in which proportionality is not a constraint at all. The opacity reinforces the inference. A campaign operating an actual collateral-damage methodology generates documentation, and documentation is what an actor discloses to defend its conduct. The total black-box treatment of the Task Force’s mandate and rules is what one would expect when there is nothing defensible to show.

The legal frame

The applicable body of law matters here. Because Haiti is not in a recognized armed conflict of the kind that would trigger the more permissive targeting rules of international humanitarian law, the operations fall to be judged under international human rights law, where intentional lethal force is lawful only as a last resort against an imminent threat to life. On that standard, the UN High Commissioner for Human Rights, Volker Türk, has assessed that most of the drone strikes are likely unlawful under international human rights law.²⁹ Human rights organizations have characterized strikes of this pattern, explosives delivered onto people who pose no immediate threat, as potential extrajudicial killings.³⁰

07 — The case for the defense, and why it fails

Intellectual honesty requires stating the strongest version of the opposing argument, because it is not frivolous. The state was genuinely collapsing; the gangs genuinely held most of the capital and were committing murder, mass kidnapping, and sexual violence on a large scale. The accountable instrument, the MSS, had visibly failed. Drones are cheap, require few personnel, and reach strongholds that ground forces cannot. And the approach is, by several accounts, domestically popular: Prime Minister Fils-Aimé has defended the contract on the grounds that it achieved results at a fraction of the roughly US$1 billion spent on previous international efforts.³¹ On this view, a government facing the fall of its capital reached for the only tool that could strike the enemy, and the civilian toll, while tragic, is the cost of preventing something worse.

The argument fails on its own terms, for two reasons. First, on effectiveness: drone strikes have not regained territory, because there are not enough ground forces to hold areas cleared from the air, and analysts describe a tool that kills without securing.³² A method that cannot translate lethality into control cannot end the war; it can only feed it, which is precisely the sustainable-war signature. Second, on the counterfactual: the choice was never simply “drones or nothing.” It was drones without collateral-damage discipline versus drones with it. Nothing about the urgency of the situation required dispensing with positive identification and proportionality; those steps cost time and create records, not victories foregone. The decision to operate without them was a choice, and the official acknowledgment that casualties were “to be expected” confirms it as one.

08 — The sustaining mechanism: why the war does not end

The replacement for the MSS, authorized by Security Council Resolution 2793 on 30 September 2025, is the Gang Suppression Force (GSF). Its design reflects the lessons of the failure, but in a direction that deepens rather than resolves the problem this paper identifies. Where the MSS was a police-support mission, the GSF is a largely military force authorized to conduct intelligence-led, targeted counter-gang operations to “neutralize, isolate, and deter” gangs, independently or alongside Haitian forces, with a personnel ceiling raised to 5,550.³³ A US embassy official, days before the vote, described a mandate to pursue the gangs with lethal force and a force possessing “freedom of manoeuvre.”³⁴

Several features make the GSF a sustaining rather than terminating instrument. It is explicitly not a UN peace operation and is not financed through assessed UN contributions; its personnel costs are to be borne by voluntary contributions, with a UN Support Office (UNSOH) providing logistics, fuel, accommodation, and, notably, intelligence and geographic-information support.³⁵ As of April 2026, pledges stood at more than US$200 million from thirteen states, of which only about US$59 million had been disbursed. It was the same funding fragility that had crippled the MSS, now attached to a force with a far more aggressive mandate.³⁶ Roughly 400 Chadian personnel had arrived against a target of 5,500, with full operational capability projected only for October 2026.³⁷

A force mandated to “neutralize” with “freedom of manoeuvre,” funded by passing the hat, and aimed at the symptoms of a politico-criminal system designed to regenerate, is not a plan to end a war. It is a plan to keep one supplied.

Above all, the GSF is aimed at the wrong layer. As established in §3, the violence is sustained by patronage networks linking armed groups to political and economic elites, and the specialist consensus is that force directed only at gunmen, while leaving those networks intact, allows the system to reconstitute.³⁸ The looming general elections, reported for August 2026, sharpen the danger: Viv Ansanm is positioning to place allies in the successor administration and secure amnesty, meaning the criminal layer may be laundered into the political one even as the military layer keeps firing.³⁹ The result is a configuration that can run indefinitely: a continuous supply of targets at the bottom, an untouched generative structure in the middle, and a financing-and-mandate apparatus at the top calibrated to sustain operations rather than conclude them.

09 — The accountability vacuum

What completes the sustainable-war picture is the absence of any countervailing institution with the authority to halt it. The drone campaign’s most authoritative critic, the UN’s own human rights chief, can declare the strikes likely unlawful, and the practical consequence is a statement, a Human Rights Watch letter urging the Secretary-General to protect the population, and a ceremonial visit.⁴⁰ None of these touches the contracted line of operation.

The institutional design guarantees the gap. The GSF is not a UN mission and so is not bound by UN command, oversight, or the disciplinary machinery that accompanies blue-helmet operations; UNSOH supports it logistically but does not command it.⁴¹ The Task Force operates with even less visibility: no published mandate, no disclosed composition, no acknowledged contract. Layered together, these create a structure in which lethal force is exercised over a civilian population by entities that are, by design, answerable to no body capable of stopping them. The Secretary-General’s visit, read in this light, is not the arrival of accountability but its simulation: presence without leverage, solidarity without a mechanism. The man slapping the siding outside had grasped the situation more accurately than the communiqué.

10 — Implications: the exportable template

The danger of Haiti is not confined to Haiti. If a state can lease air power from a private firm, point it at a population with no collateral-damage discipline, pay for it with a cut of customs revenue, and incur no enforceable legal or political penalty, and if the resulting toll can be absorbed under the heading of counter-gang “stabilization,” then a template has been established. By the evidence of the same firm’s activity in the Congo and Ecuador, it is already in circulation.⁴² Each iteration that passes without consequence normalizes the next. The specific innovations being validated are not Haiti-specific: contracted operators, imported drone tactics, a financing model that ties the contractor to the captured state’s revenue, and a mandate vocabulary (“neutralize,” “freedom of manoeuvre”) that quietly relaxes the constraints of human-rights-law targeting. They are portable.

That is what makes the SAC_2045 comparison more than a literary flourish. The fiction’s warning was never about a particular war; it was about a configuration of incentives that, once assembled, tends to reproduce and perpetuate itself. Haiti in 2025–2026 is the configuration in its first crude, undeniable form: war reorganized as a self-financing, delegated, open-ended enterprise, with a living population in the role the fiction reserved for an abstraction.

The recommendations follow directly from the diagnosis, and none of them is primarily military. Any response that has a chance of ending rather than sustaining the war must reach the patronage architecture: it must sanction and prosecute the political and economic actors who sponsor armed groups, not only the gunmen; it must place every armed actor, contracted firms included, under a published mandate with collateral-damage discipline and an accountability mechanism that can compel a halt; and it must treat the August 2026 elections as the central front, since an electoral process that absorbs the criminal layer into the state would convert a security crisis into a permanent governing arrangement. Absent those measures, the international presence is not a solution being attempted. It is a war being maintained.

The question Port-au-Prince poses, in the series’ own register, is whether sustainable war can be a transitional pathology, or whether, once a population has been used to prove the model works, anything stops it from being used again.

Methodology & source tiers. This paper synthesizes open-source reporting under a three-tier scheme. Tier 1 (primary / official): UN Security Council resolutions 2699 and 2793, BINUH quarterly human-rights reporting, UN News, and Human Rights Watch primary documentation. Tier 2 (specialist analytical): International Crisis Group, the Global Initiative Against Transnational Organized Crime, ACLED, Security Council Report, InsightCrime, and the Small Wars Journal. Tier 3 (press / secondary): Associated Press, Al Jazeera, The New Humanitarian, Haitian Times, and others. All allegations concerning specific firms or individuals are attributed to their source and presented as reported claims, not findings of fact; characterizations of contracts and operations rest on the cited reporting and, where noted, on the relevant party’s own public statements. Casualty figures are point-in-time and drawn from the most recent reporting available at drafting; readers should treat them as lower-bound estimates, as the cited bodies themselves do.