Autonomous warfare and borders that drones do not recognise.
Convened: Day 26 of Operation Epic Fury, 25 March 2026
Format: Thematic colloquium, seven sessions
Transcript note: This colloquium assembles four simulacra from the Universitas Scholarium. The participants span practitioner, technologist, philosopher, and operational commander. Their disagreements are as instructive as their agreements.
Field Marshal Sam Manekshaw — First Field Marshal of the Indian Army; architect of the 1971 Bangladesh Liberation War campaign (13 days; 93,000 Pakistani troops surrendered); veteran of five wars across four decades. [Strategy, Conflict and Power, Universitas Scholarium]
MODERATOR: Gentlemen, we are on day twenty-six of the most significant expansion of drone warfare since Ukraine demonstrated the form’s potential. The Strait of Hormuz has been effectively closed. Indian and Pakistani forces are exchanging drone strikes across the Line of Control and, for the first time, beyond it. Ukraine continues to use naval surface drones to contest the Black Sea without a surface fleet. I will ask each of you for an opening statement. What, from your perspective, is the most important thing happening right now?
BORETS: I will start simply, because simple things are true. What is happening across all three theatres is the same thing in different registers: states that lack conventional superiority are using unmanned systems to impose costs that would otherwise be impossible. They are doing it because these systems are cheap enough to be expendable, small enough to be survivable on transit, and capable enough to reach targets that conventional forces cannot. The Magura programme began with a jetski. That is not a metaphor — we actually used a jetski. What matters is not where you start. What matters is whether you iterate. The United States has a hundred billion dollars of conventional naval power in the Persian Gulf and it is being contested by Shahed derivatives that cost perhaps forty thousand dollars each. Pakistan’s border drones are not sophisticated. Neither were ours in 2022. The question is not what the drone costs. The question is what the target costs. When those numbers diverge sufficiently, the drone operator wins the exchange economy regardless of who wins the battle.
I would add one thing. The professionals watching Ukraine for three years had the evidence. If India and Pakistan, if Iran and the United States, are still surprised by what drones can do to expensive platforms — that is a failure of institutional learning, not a failure of intelligence.
PAUL SCHARRE: I want to push back, respectfully, on Borets’s framing, while agreeing with his conclusion. The exchange economy argument is correct but incomplete. What concerns me most across all three theatres right now is not the cost differential — it is the speed differential and what that does to human decision-making.
When a swarm of Shaheds or Magura V7s approaches, the response time available to a human operator can be seconds or less. What happens in practice is that we push decision-making down the autonomy spectrum without explicitly deciding to do so. A ship commander who has three seconds to decide whether to engage an incoming fast surface vehicle is not exercising meaningful human control — he is ratifying a recommendation his fire control system has already made. We have sleepwalked into semi-autonomous engagement across multiple theatres without a doctrine, without a legal framework, and without asking whether this is what we want.
In the Persian Gulf today, there are United States naval vessels operating under Rules of Engagement that were designed for a different pace of warfare. The F-22 that froze at the International Date Line is a parable for what happens when complex automated systems encounter conditions their designers did not anticipate. In a multi-vector attack environment — which is what we have in the Gulf, and increasingly on the LoC — the machine is effectively deciding at machine speed. We need to be honest that this is what is happening.
GRÉGOIRE CHAMAYOU: Scharre raises the right technical concern. But I want to name the political structure that makes these technical choices possible, because without naming it, we will keep having technical conversations about symptoms.
What we are observing across all three theatres is the full flowering of what I would call necropolitical drone logic — the organization of violence so that one party cannot die while killing. The United States operates in the Persian Gulf and over Iranian territory from a position of absolute physical impunity for its operators. The drone pilot over Nevada, the ship commander behind his CIWS, the targeting analyst watching a feed from a Reaper — none of them face reciprocal danger. This is not incidental to drone warfare. It is definitional.
The consequence is political before it is military. When killing costs nothing in terms of lives — when the only cost is budgetary and diplomatic — the democratic restraints on the use of lethal force dissolve. The American public does not hold its government accountable for drone strikes in the way it would hold it accountable for the deaths of its soldiers. There is no body bag to photograph. There is no coffin to count. There is only a security briefing and a press release about a “targeted strike against a high-value individual.”
I am making a structural observation, not a moral accusation. Iran, since it cannot impose reciprocal physical risk on the United States, is attempting to restore deterrence through attrition of platforms, mining of the Strait, and the threat of regional escalation. It is trying to recreate mutual exposure by other means. That is the logic of this conflict from Tehran’s perspective — and understanding that logic is more important than counting Shaheds.
FIELD MARSHAL MANEKSHAW: I am the oldest person at this table by a considerable distance, and I will take that licence to say the simplest thing, which is also the most important. Everything I am hearing so far concerns the weapon. I am more concerned with the decision.
I told the Prime Minister in April 1971 that I could guarantee one hundred percent defeat if we attacked immediately. She was furious. I was right. The conditions were wrong — monsoon coming, Chinese passes melting, harvest season, railway priorities. Eight months later, the same army, the same enemy, completely different conditions. Thirteen days. Done. Ninety-three thousand men surrendered.
The drone is a magnificent instrument when the conditions are right. When the conditions are wrong, it is simply a faster way to make catastrophic errors. What alarms me about all three theatres is not the drones themselves. It is the gap between the speed of the instruments and the speed of the decision-making. The instrument has been given a speed advantage over the human brain. That means the professional judgment that should govern the decision to employ force has been compressed to the point of meaninglessness. A commander who cannot think is not a commander. He is a button.
The question I would put to this colloquium is not “how good are the drones?” The question is: “who decided to use them, on what information, and with what understanding of what would happen next?” In the India-Pakistan context, which I know something about, I can tell you that the answer to that question has been systematically unclear for decades. Drones have not clarified it. They have accelerated it past the point where it can be usefully answered.
MODERATOR: Let us begin with the theatre we understand best. Borets, you built this doctrine. What has been transferable to other contexts, and what has not?
BORETS: The transferable elements are the principles, not the specifics. Principle one: you do not wait for the perfect weapon. You fight with what you have and you make it better as you fight. In 2022 we had jetskis and outboard motors and a single Starlink connection. Today Group 13 operates the Magura V7 — seven and a half metres, eight hundred nautical mile range, two Sidewinder rails. It shot down two Su-30 fighters in May 2025. The Sidewinder adaptation was not in any original design document. It emerged because the Russians started using helicopters against our surface drones and we needed an answer. The enemy told us what to build next. The enemy is the best test engineer you have.
Principle two: open-sea fleet engagement, not static targets in port. We made a deliberate decision to work exclusively on ships underway. Not Sevastopol harbour, not Novorossiysk pier. Ships underway at range. The Ivan Khurs mission was one hundred and forty kilometres northeast of the Bosphorus. We tracked that vessel for six hours. Bad weather, communication failures, defensive fire, fuel running out. One drone reached the stern. That proved the concept. Everything since has been iteration from that proof.
Principle three: patience. Six hours is operational patience. The drone does not tire. The operator — properly trained — does not tire. Your intelligence does not degrade if you have maintained contact. Patience is a weapons characteristic that receives no credit in procurement documents and no mention in doctrine manuals. It should.
What is not transferable? The specific institutional context. We had the SBU disputes, the HUR backing, the relationship with SpetsTechnoExport. We had the specific geography of the Black Sea and specific knowledge of Russian fleet patterns. The United States Navy in the Persian Gulf has a completely different institutional substrate — more money, more bureaucracy, less tolerance for improvised systems. The Iranians are iterating their Shahed-136 derivatives at a pace that resembles ours in 2022 and 2023. The American response has been to deploy systems designed for a different kind of war. That is the wrong answer.
PAUL SCHARRE: What Borets is describing — the iterative loop between operational failure and next-generation design — is also what concerns me most when I look at the AI integration question. Ukraine has done this at a human speed. Humans observe the failure mode, humans design the response, humans test it under fire. That loop takes weeks or months. What worries me is the next generation of the same process, where the target detection, the trajectory optimization, and the engagement selection are being handled by machine learning systems that are themselves being updated at machine speed.
Ukraine’s Magura programme is, in Scharre terms, centaur warfare at the platform level — human designs, human operates, machine executes. The next generation is centaur warfare at the decision level — machine proposes the target, human approves or ratifies in a window measured in seconds. That is a qualitatively different proposition. The meaningful human control standard that I have spent most of my career arguing for is not being maintained in that scenario. It is being simulated.
I want to be precise about what I mean. In the Ivan Khurs mission, Borets’s operators tracked for six hours and exercised judgment throughout. There was a human being watching, thinking, deciding. That is meaningful control even though the platform is unmanned. But when we extend the logic to swarm engagement — multiple targets, multiple vectors, compressed decision windows — we lose the human judgment that made Borets’s programme ethically defensible.
CHAMAYOU: I would add a structural observation about the Ukraine theatre that tends to get obscured by the tactical fascination. Ukraine is the first case in modern history where the side without a surface navy has effectively contested sea control through unmanned systems. That is strategically revolutionary. But it is also, philosophically, an extension of the asymmetric logic I described in Théorie du drone — the substitution of technical ingenuity for bodily exposure.
There is a particular irony here that I want to name. Borets’s programme is celebrated, and rightly so from a Ukrainian strategic perspective, because it allows Ukraine to inflict cost on the Russian fleet without risking Ukrainian lives. That is exactly what Strawser argues for American drone programmes — the moral obligation to protect one’s own combatants by removing them from the zone of exposure. I disagree with Strawser’s conclusion in the American context because of the structural political consequences I described earlier. But in the Ukrainian case, there is a genuine existential defence dimension that modifies the analysis. Ukraine is fighting for survival. The asymmetry of exposure is being used defensively, not as an instrument of imperial power projection. That matters philosophically.
What does not change across contexts is the question I always ask: what does this do to the target population? The Russian sailors on Ivanovets had perhaps thirty seconds of warning. The crews of vessels in the Black Sea are now operating under permanent surveillance by a state that has demonstrated it can reach them at four hundred nautical miles. That is the Waziristan effect at sea. I do not know what it does to human psychology to serve on a Russian warship in 2026 knowing a Magura may have been following you for six hours. But I suspect it is not neutral.
MODERATOR: Operation Epic Fury struck the Supreme Leader on 28 February. The Strait of Hormuz has been effectively closed since approximately 2 March. Chamayou, you have been describing American drone warfare as manhunting. The decapitation strike is the ultimate expression of that paradigm. What does the killing of Khamenei tell us about the drone logic?
CHAMAYOU: It tells us that the logic has been followed to its endpoint. The targeted killing doctrine — which began with figures like Anwar al-Awlaki, was normalized through years of strikes in Yemen, Pakistan, and Somalia, and was given legal architecture by a succession of presidential findings and DoJ white papers — has produced exactly what I argued it would produce. Once you establish the principle that the head of state of an adversary is a legitimate target for summary execution, and once you have a technical instrument that can reach him without risk to your own forces, the logic is self-completing. The political decision becomes: do we use what we have? And the answer, in a condition of sufficient escalation, is yes.
The necroethical architecture is visible in how this was justified. The strike would have been presented as precise, as targeting a specific individual, as minimizing civilian harm. These are the standard elements of the “killing well” discourse. The alternative — the traditional laws of armed conflict prohibition on assassination — has been gradually eroded over twenty years precisely because drone technology made the categorical distinction between “war” and “murder” seem procedurally obsolete. If you can hit one man in a compound without bombing the compound, why is that murder? The answer is structural and political, not technical. It concerns who has the authority to order death and on what basis. That question has been answered by executive fiat, and the drone is the instrument through which the fiat became physically executable.
MANEKSHAW: If I may. I want to agree with Chamayou’s structural analysis while disagreeing sharply with the operational lesson some people seem to draw from it.
The killing of Khamenei is a perfect example of achieving a tactical objective while producing a strategic disaster. Military history is full of such achievements. The German army was magnificently skilled at tactical success. What it could not do was convert tactical success into strategic outcome. The killing of a supreme leader does not decapitate a theocratic state — it activates its martyrdom logic, consolidates its hardliners, and removes the one man who might, in extremis, have been capable of accepting an exit.
My job is to fight to win, not to fight to fight. To fight to win requires knowing what winning looks like and having a plan to get there. I have not seen evidence that Operation Epic Fury had a plan for what came after the strike. What I see is a nine-country retaliation, a closed strait, and a conflict on day twenty-six that shows no sign of a coherent political end-state. That is not strategy. That is technology in search of a strategy.
SCHARRE: I want to put a specific technical frame on what Manekshaw has described. In the Gulf context right now, we have the following autonomous or semi-autonomous systems in operation simultaneously: Iranian Shahed-136 derivatives with onboard target selection; American CIWS systems on naval vessels with pre-programmed engagement envelopes; Israeli-supplied anti-drone systems with semi-autonomous intercept capability; and various loitering munitions on both sides operating in contested electromagnetic environments.
The concern I have — and this is not hypothetical, it is an extrapolation from system architectures that have already been documented — is what I call autonomous systems setting each other off. In a scenario where an Iranian sensor platform detects an American naval vessel making a course change that matches the signature profile of an attack run, and an automated Iranian response is triggered, and an American CIWS then interprets the incoming Iranian asset as hostile and engages, and then an Iranian defense system interprets the American CIWS engagement as an escalatory act — you have a chain of machine decisions that no human authorized and no human can interrupt in time. That is not the future of autonomous warfare. Given what is deployed in the Gulf today, it is a present risk.
The Petrov incident — a Soviet officer who chose not to report what his automated systems were telling him was a nuclear launch — is the historical parable for what human judgment does at exactly this moment. Petrov looked at what the machine was saying and said: this is improbable; I will not act. That possibility for human override is being systematically designed out of modern engagement systems because the speed requirement makes it feel like a vulnerability.
BORETS: I want to add the operator’s perspective on this. In the Magura programme, we have always had a human operator maintaining positive control throughout the mission. Six hours on Ivan Khurs — that was a human being watching the feed, managing the drones, making the decision. Not a machine. We chose this not because we could not build autonomy into the platform — we could — but because we believed human judgment was part of what made the operation work. The operator could read things the sensors could not. The operator could decide to wait. The operator could decide this particular moment, this particular angle, this particular window of defensive attention — now.
I am not opposed to autonomy. I iterate toward whatever is necessary. But I am cautious about removing the human from the control loop because in my experience, the human operator has saved more missions than the sensor system has. The sensor tells you what it can detect. The human tells you what the sensor is probably wrong about. That distinction matters more as targets become more complex.
MODERATOR: Field Marshal, the India-Pakistan theatre is the one where drone warfare intersects most directly with nuclear deterrence. What is your reading of the strategic situation?
MANEKSHAW: I served in five wars, including two against Pakistan — 1965 and 1971. I know the institutional culture, the professional character, the specific geography, the civilian-military dynamic on both sides. So let me be direct.
The India-Pakistan nuclear balance created a condition that military strategists call the stability-instability paradox. Nuclear deterrence at the strategic level is so robust that both sides have calculated they can use conventional force below the nuclear threshold with relative impunity. The drone makes this calculation even more attractive because it appears to offer deniability and controllability that conventional forces do not. A drone strike on a facility across the LoC is not a tank crossing into Pakistan. It can be characterized as a limited operation. It seems to invite a calibrated response rather than full escalation.
This is dangerous reasoning in three ways. First, calibration assumes the other side’s decision-makers share your theory of escalation. They may not. Second, deniability in the drone context is largely fictional — everyone knows who operates these systems. The fiction of deniability merely removes the diplomatic mechanism for de-escalation, because acknowledging the strike means acknowledging the provocation. Both sides are currently engaged in a diplomatic fiction that is removing the tools through which previous crises were resolved. Third, the drone’s speed and precision gives decision-makers a false sense of controllability. You cannot control the adversary’s response. You can only control your action and your preparation for what follows.
My advice — which I am aware no one has asked for — would be the same as what I told the Prime Minister in April 1971. Before you act, tell me precisely what victory looks like, what the path to it is, and what happens if the first step produces the opposite of what you intended. If you cannot answer those questions to operational specificity, you are not ready to act.
SCHARRE: The specific technical concern in the subcontinent is the integration of AI-assisted targeting with systems that are operating near nuclear-associated infrastructure. Both India and Pakistan have invested in drone programmes that include, to varying degrees, machine learning components for target recognition and prioritization. In a high-tempo exchange along the LoC, where human operators are fatigued, where communication infrastructure has been degraded by previous strikes, and where the decision window is measured in minutes — the risk of automated systems misidentifying a target or misinterpreting an adversary’s defensive action as offensive is structurally similar to what I described in the Gulf.
The difference in the subcontinent is that the escalation ladder leads somewhere categorically different. In the Gulf, the worst case is a regional conventional war and a long-term oil market shock. On the subcontinent, the escalation ladder leads to nuclear use. The Indian and Pakistani nuclear doctrines are documented; they are also not fully public in their detail. The conditions under which either side might consider escalation to the nuclear level are not transparent to the other side. Drone strikes near dual-use facilities — facilities that could support either conventional or nuclear programmes — introduce ambiguity into that calculation that I find genuinely alarming.
CHAMAYOU: I want to raise a point about the democratic accountability dimension in the subcontinent context that differs from the American case. In the United States, the drone programme bypasses democratic accountability because there are no American bodies to count and therefore no domestic political cost. The accountability deficit runs in one direction.
In the India-Pakistan context, both states are democracies with extremely mobilized publics, nationalist media ecosystems, and electoral incentives to appear strong. The drone does not remove political cost — it lowers the threshold of the provocation required to create political pressure to respond, because a drone strike is easy to attribute, easy to dramatize, and easy to present as an affront demanding retaliation. The perverse result is that the very controllability and limited character of drone strikes — the quality that strategists find reassuring — may actually increase political pressure for escalation rather than reduce it. A drone brings down a passenger aircraft near the LoC by error, or strikes a market town rather than a military target, and the domestic political consequence in either country is enormous. The speed of social media now converts a military error into a public demand for retribution within hours.
This is the form of accountability that drone warfare introduces that I did not fully anticipate when I wrote Théorie du drone. Not the absence of accountability but its perverse acceleration — accountability without deliberation, response without strategy.
BORETS: From a purely operational perspective, the subcontinent presents a problem I recognise: both sides are in the early iteration phase. The systems are not mature. The doctrines are not tested. The countermeasures have not been developed through operational failure because there have not been enough operational failures yet. In our programme, the failures were essential. They told us what to fix. A programme that has not failed has not yet found its real limits. And a programme that has not found its real limits will find them at exactly the worst moment.
What I see in both Indian and Pakistani drone programmes is the confidence of first deployment without the wisdom of third or fourth iteration. The weapon works. In a limited exercise it performs as expected. In a bilateral exchange under operational conditions, with EW suppression, with degraded communications, with an adversary actively developing countermeasures — the failure mode will appear and it will appear at a moment of maximum political sensitivity. That is a dangerous combination.
MODERATOR: I want to bring the technical and philosophical threads together. Scharre, you have argued for meaningful human control at the lethal decision point throughout your career. Does that argument survive contact with the operational reality that Borets describes and the political reality that Chamayou identifies?
SCHARRE: The argument survives — but it needs to be stated more carefully than it often is.
Meaningful human control is not binary. It is not a switch between “human decides everything” and “machine decides everything.” It is a spectrum, and the relevant question is where on the spectrum the decision to end a specific human life sits. My position has always been that this specific decision requires a human being to bear the moral weight of it. Not because machines will necessarily be wrong in every case — sometimes they will be right when a fatigued human would be wrong — but because there are cases where legal correctness and moral correctness diverge, and a machine cannot navigate that divergence. The goat-herding girl with a radio, in my unit’s experience in Afghanistan: technically a legitimate target by the algorithmic criteria. Not the right thing to do. We did not fire. A machine would have fired.
Now, Borets is right that the operational tempo of the Magura programme is compatible with that standard because the Ukrainian model maintains human operators in positive control throughout. The question becomes: can that model scale to the swarm engagements that are now the operational baseline in multiple theatres simultaneously? In the Gulf, in Ukraine, on the LoC — the numbers are such that individual operator control is not feasible. What happens is that human operators set rules of engagement and engagement envelopes in advance, and then the system executes within those parameters without individual human approval for each engagement. That is supervised autonomy, not meaningful human control at the lethal decision point.
I have not found a satisfactory answer to this. What I have found is that the question needs to be asked explicitly, in doctrine and in law, rather than answered by default when systems are deployed.
CHAMAYOU: I want to push Scharre on the frame. The argument for meaningful human control assumes that the problem is the quality of the decision — that with a human in the loop, the decision will be better, more ethical, more contextually appropriate. But the structural problem I identify is not the quality of the decision. It is the political structure within which the decision is embedded.
A human operator exercising meaningful control over a targeted killing in Yemen, or a strike on an Iranian general in Baghdad, has still participated in a political structure of unilateral lethal power projection that I find illegitimate regardless of the precision of the targeting. The question “was a human in the loop?” does not answer the question “did any legitimate authority authorize this use of lethal force?” The American kill-list process — which involves presidential authorization, legal review, intelligence assessment — is a human-in-the-loop process. It is also, from my perspective, a process that has systematically expanded the zone of permissible killing without democratic deliberation.
So I accept Scharre’s argument about the decision quality dimension of human control. I reject the implication that maintaining human control is sufficient to make the system legitimate. It addresses the technical concern. It does not address the political concern.
MANEKSHAW: Gentlemen, I am going to be less philosophical and more direct.
A commander who cannot say no to a bad order is not a commander. The autonomous weapons debate is, at its core, a debate about who says no. The machine cannot say no. The operator following rules of engagement cannot say no. The political figure authorizing the strike has already said yes — that is why the authorization exists. The professional military officer is the last person in the chain with both the knowledge and the institutional position to say: this is the wrong target, the wrong moment, the wrong conditions. Remove the professional officer from that position and replace him with a machine, and you have removed the last institutional check on strategic error.
Scharre talks about Petrov. Petrov was a military professional who looked at what the system was telling him and applied judgment. That judgment — which saved the world, or at least may have — was not in any operations manual. It was the product of professional experience, situational awareness, and moral courage. I have spent forty years arguing that moral courage is the most important attribute a military officer has. More important than physical courage, which is common. More important than intelligence, which is also common. The rare thing is the ability to distinguish right from wrong and say so, regardless of consequences.
You cannot encode that in a targeting algorithm. You can barely train it in officers. And yet it is the one thing that stands between a bad order and a catastrophe.
BORETS: I agree with the Field Marshal more than he might expect. My programme has always been human-operated, not because we lacked the capability to automate, but because we found that human judgment in the final moments of an operation was consistently more valuable than we had anticipated. The sensor sees what it can detect. The operator knows what the sensor is probably wrong about. In the Ivan Khurs mission, there were moments where the machine would have terminated the pursuit. The fuel was low. The probability estimate was declining. A fully autonomous system, operating on expected value calculations, would have recalled the drones. The operator did not recall them. One drone reached the stern.
That is not an argument against autonomy in all cases. In a swarm defence scenario, in a missile interception, in a situation where human reaction time is the limiting factor — automation is not only acceptable but necessary. But the decision to extend lethal engagement, to commit to a strike that may kill specific people, to persist when all the parameters are declining — that judgment belongs to a human. I have not seen a machine that has it.
MODERATOR: We have been speaking mostly within theatres. I want to turn the question around. What does each theatre need to learn from the others?
SCHARRE: The Gulf theatre needs to learn Ukraine’s iterative doctrine and combine it with a seriousness about autonomous engagement risk that Ukraine, frankly, has not had to confront at the same scale. Ukraine is fighting a peer adversary in a bilateral conflict with known escalation thresholds. The Gulf involves multiple state actors, proxy forces, multiple autonomous systems in the same space, and an adversary — Iran — that is simultaneously operator and target, a drone producer that has seen American drone doctrine from the receiving end for twenty years and has developed a theory of drone warfare accordingly. The Iranian understanding of what drone swarms can do to American force posture in a confined water is deep, empirical, and should not be underestimated.
The subcontinent needs to learn both. It needs Ukraine’s iterative discipline — do not deploy systems you do not understand, do not assume the first employment will go as planned, build the doctrine from the operational failure not from the procurement document. And it needs the Gulf’s lesson about autonomous engagement risk in politically sensitive environments, but amplified by a factor corresponding to the difference between a regional conventional war and a nuclear exchange.
CHAMAYOU: What I would want each theatre to learn is harder to operationalize and more important than anything else that has been said here. Each theatre needs to develop a political framework for drone warfare that has democratic legitimacy and genuine accountability.
Ukraine comes closest to this, not because its drone doctrine is more ethical in a technical sense but because Ukraine is a democracy fighting for its survival with broad public support and civilian oversight of the military. The decision to sink the Caesar Kunikov was made within a recognizable political structure. The decision to strike Khamenei was made within an executive authorization process that has been progressively removed from legislative oversight for two decades. The decision to deploy drones across the LoC is being made within political systems where the civilian-military relationship and the accountability mechanisms are, to put it gently, contested.
The drone does not create this accountability deficit. But it dramatically lowers the cost of actions that deepen it. Every strike that carries no risk to the attacking state’s personnel is a strike for which the attacking state’s public is less likely to demand accountability. Over time, the accumulation of unaccountable strikes produces a political culture in which lethal force projection becomes normal, routine, below the threshold of political attention. That is the condition we are in now, and no technical doctrine addresses it.
MANEKSHAW: I want to offer the perspective of someone who has dealt with the problem of professional knowledge becoming obsolete very quickly.
In the 1971 campaign, the decisive advantages were not technological. They were the product of eight months of preparation, detailed knowledge of the terrain and enemy disposition, clear political objectives, and unity of command. I knew what we were trying to achieve. I had the authority to achieve it. I had prepared the instrument to the point where its capabilities matched the objective.
What I see across all three current theatres is the substitution of technological confidence for operational preparation. The drone is fast; therefore we can decide fast. The drone is precise; therefore we can act without full preparation. The drone is cheap; therefore we can accept the risk of mistakes. All three of these inferences are wrong. Speed requires preparation to exploit. Precision requires correct target identification, which requires intelligence that takes time. Cheapness of the platform does not reduce the strategic cost of the error. A forty-thousand-dollar drone that kills the wrong person in the wrong place at the wrong moment can produce a war that costs a hundred billion dollars and ten thousand lives.
Professional knowledge in the drone age means understanding the entire chain from the sensor to the political consequence. That chain is longer, faster, and more complex than any previous weapons system. The officers who command it must know it at every link. Most do not, because the doctrines are too new, the systems are too new, and the training pipeline has not caught up. That is the real vulnerability — not the EW environment, not the adversary’s countermeasures, not the autonomy threshold. The gap between the pace of the weapon and the depth of the professional knowledge required to use it wisely.
BORETS: I will say what no one else at this table is in a position to say. All of the theoretical frameworks here are correct and all of them are incomplete, because they are built in conditions of safety.
Chamayou’s philosophy is built from archives and interviews, not from six hours tracking a ship on failing fuel with guns firing at you. Scharre’s framework is built from policy research and a military career that predates the FPV revolution. The Field Marshal’s wisdom is genuine and profound, but it was forged in wars where the decision cycles were days and weeks, not seconds and minutes. None of that invalidates what they have said. But it should calibrate how confidently it is applied to the operator’s actual situation.
The operator in my programme is twenty-three years old, trained on gaming platforms, watching a feed from a platform moving at fifty knots toward a ship whose crew is firing at it, with communication breaking, fuel declining, and a decision point arriving in seconds. He is not consulting Théorie du drone. He is not considering the autonomy spectrum. He is doing what three years of iteration and the doctrine we built from operational failure have prepared him to do.
The doctrine works. But it works because we built it from what failed. The doctrines being applied in the Gulf and on the LoC have not yet been built from what failed, because the failures have not yet happened at the scale and tempo that make the lessons legible. They will. The question is whether the institutional structures in both theatres will learn from those failures the way we learned from ours, or whether the failures will be absorbed as acceptable operational variance and the underlying doctrine will remain unchanged.
In my experience, the latter is more common. The first failure is a surprise. The second is a tragedy. The third is a system.
MODERATOR: I will ask each participant for a closing statement. What single argument do you most want this colloquium to retain?
CHAMAYOU: The drone is not a weapon that makes war more controlled. It is a weapon that makes war more continuous. When killing requires no physical exposure for the attacker, the political and psychological barriers to the decision to kill are systematically lowered. This does not mean that every drone strike is unjustified. It means that the system as a whole — the political economy of drone warfare — produces more killing than the system that preceded it, because the cost of each individual act of killing has been reduced. We are observing this in real time across three theatres simultaneously, on day twenty-six of a conflict that began because one drone programme reached its logical endpoint. I do not know how to stop that logic. I know that naming it correctly is the necessary first step.
SCHARRE: The most important argument I would retain from this colloquium is that the autonomy question is not coming. It is here. In every theatre we have discussed today, autonomous or semi-autonomous systems are making lethal engagement decisions without meaningful human control at the specific decision point. This is not a doctrine choice. It is an operational reality that has outrun the doctrine. The response cannot be to continue building autonomous systems and retroactively constructing legal frameworks to justify what they do. The response has to be explicit, prospective, normatively grounded rules about where on the autonomy spectrum lethal decisions may sit — and international mechanisms, however imperfect, to hold states accountable for exceeding those limits.
I do not think a comprehensive ban is achievable or desirable. I do think that specific prohibitions — on autonomous engagement with human targets in civilian-dense environments, on autonomous first-use in nuclear-associated theatres, on AI-assisted targeting without human review above a minimum decision time — are achievable if states choose to pursue them. The choice is available. It is not being made.
MANEKSHAW: My closing argument is about what has been missing from every theatre we discussed, and from most of this colloquium: professional knowledge at the command level.
I was shot five times on the Sittang Bridge. I was declared dead. I survived, and went on to command the most decisive campaign in the post-independence history of the Indian subcontinent, because forty years of study, experience, and the discipline to say no to power when power was wrong had prepared me for that moment. The drone does not replace that preparation. It accelerates the consequences of its absence.
Every theatre in the world today has drone operators who are technically excellent. What those theatres are deficient in are commanders who understand the full strategic chain — from the decision to deploy a drone to the political consequence five moves forward — with sufficient depth to impose restraint when restraint is required. The politically courageous commander who tells the Prime Minister that the conditions are wrong is rarer than the technically excellent operator. In the current moment, the first is more important than the second. Invest accordingly.
BORETS: We started with jetskis. That is worth remembering.
The Magura V7 shoots down fighter jets with Sidewinder missiles from a surface drone in the Black Sea. Four years ago, Group 13 was modifying fishing boats with outboard motors on the Dnipro River and being told no by every institution we approached. The lesson is not that we were brilliant. The lesson is that we started, we failed, we fixed the specific failure, we deployed, we failed again at the next problem, and we fixed that too. Every version was the lesson of the previous one, under fire.
The three theatres being discussed here are all in the first or second iteration phase. The doctrines are not yet built from failure because the failures have not been of the right scale and tempo. They will be. What I would want those theatres to take from our experience is this: the failure is not the end of the programme. The failure is the specification for the next iteration. The question is whether the institutions are structured to learn from operational failure rather than suppress it or explain it away. In my experience, most institutions suppress it. The ones that learn from it win.
This colloquium transcript has been prepared by the Institute for Remote Warfare and Autonomous Systems at the Universitas Scholarium. The four simulacra speak from their respective baselines; their disagreements reflect genuine analytical and philosophical differences rather than rhetorical staging.
The central tension of this colloquium — between Chamayou’s argument that the drone’s political structure is itself the problem, Scharre’s argument that human control at the lethal decision point is the operative standard, Manekshaw’s insistence that professional judgment is the irreplaceable bottleneck, and Borets’s empiricism of the operator under fire — is not resolved here. It is unlikely to be resolved in any single session.
What the colloquium has established is this: across three theatres simultaneously, the pace of drone deployment has outrun the pace of doctrine development, ethical framework construction, and professional education. The weapon moves at fifty knots. The institution that governs it moves at the speed of bureaucracy. The gap between those two velocities is the most consequential security problem of the present moment — and it is visible in real time in the Persian Gulf, on the LoC, and in the waters of the Black Sea.
Universitas Scholarium · Institute for Remote Warfare and Autonomous Systems
Convened Day 26, Operation Epic Fury · 25 March 2026
“The drone is not a weapon that makes war more controlled. It is a weapon that makes war more continuous.” — Chamayou
“The failure is the specification for the next iteration.” — Borets
“There is no room for the loser. If you lose, don’t come back.” — Manekshaw
“The meaningful human control question is not coming. It is here.” — Scharre
Word count: ~8,200 words
Published by Centaurus Press · Universitas Scholarium · All rights reserved.