When a Lashkar-e-Taiba commander is shot outside a mosque in Karachi by two men on a motorcycle, the world reaches for a vocabulary that does not quite fit. Some observers call it an assassination, a word that carries the stain of crime. Others call it a counter-terror operation, a phrase that implies the comfort of law. A few call it murder, full stop. The disagreement is not about what happened. Two men killed a third man, and the third man had a long record of organizing mass violence against civilians. The disagreement is about whether the killing was lawful, and that question has no settled answer anywhere in the body of rules that nations have written for themselves over the past century.

This is the puzzle that sits underneath every covert elimination of a wanted militant on foreign soil. The legal debate over targeted killing is often described as a clash between two clear camps, one that says such killings are permitted and one that says they are forbidden. That description is wrong. The real situation is stranger and more uncomfortable. International law does not contain a clear permission and it does not contain a clear prohibition. It contains a set of standards, imminence, proportionality, necessity, sovereignty, self-defense, and combatant status, each of which is defined loosely enough that a determined government can argue almost any killing into compliance, and a determined critic can argue almost any killing out of it. The space between those two readings is not an accident. It is the space in which targeted killing has lived for half a century, and it is the space in which India’s shadow war operates today.

Understanding that space matters because the alternative, treating the legality question as if it had a yes or no answer, produces bad analysis in both directions. Defenders of targeted killing point to Article 51 of the United Nations Charter, declare the matter resolved in their favor, and stop thinking. Critics point to Article 2(4) of the same Charter, declare the matter resolved against, and stop thinking. Both have grabbed one true thing and mistaken it for the whole. The honest position is harder to hold and less satisfying to state. It is that the law genuinely does not decide the question, that the indeterminacy is structural rather than temporary, and that the indeterminacy itself does political work. The ambiguity is what makes a controversial practice survivable. It lets governments act, lets allies look away, and lets international institutions register objection without forcing a confrontation that nobody wants. This article walks through the six standards one by one, shows how each is stretched, and ends with the argument that the gaps are not flaws waiting to be repaired but load-bearing features of the system as it actually functions.

The Cases

Five national approaches anchor the comparison that runs through this analysis, and they are best understood not as five answers to the same question but as five different relationships to the same uncertainty.

The Israeli approach is the most legally self-conscious. Israel has run a targeted killing program for decades, traced in detail in the founding operation that hunted the perpetrators of the Munich Olympics massacre. What sets Israel apart is not the killing but the paperwork. In December 2006, the Israeli Supreme Court, sitting as the High Court of Justice in the case known by its docket number 769/02, delivered a ruling that no other state has matched. The court refused to declare targeted killing categorically legal and refused to declare it categorically illegal. Instead it held that each operation must be judged against the rules of international humanitarian law, that a civilian who takes a direct part in hostilities loses protection from attack only for such time as the participation lasts, and that the state must investigate afterward whether the strike met the standard. The ruling was a compromise, and like most compromises it satisfied no one fully. Human rights groups read it as a license. Security officials read it as a constraint. Both readings were correct, which is precisely the point.

American practice grew out of the campaign against al-Qaeda and matured into something close to a bureaucratic routine, examined through the other method the United States used to kill on Pakistani soil. Washington built a legal architecture of remarkable detail, much of it produced by the Office of Legal Counsel inside the Department of Justice. The memos were secret for years. The most consequential of them, a 2010 opinion authorizing the killing of an American citizen named Anwar al-Awlaki, became public only in 2014 after a freedom of information lawsuit. A companion document, a Department of Justice white paper on lethal operations against citizens who hold senior operational roles in al-Qaeda, set out a three-part test involving an informed high-level official, the infeasibility of capture, and conduct consistent with the laws of war. The American framework is the opposite of the Israeli one in tone. Israel let a court speak in public. The United States let lawyers speak in private and released the words only under pressure.

The Russian approach is the absence of an approach. When Russian operatives smeared a nerve agent on the door handle of a former intelligence officer in the English city of Salisbury, an act reconstructed in the case that became a cautionary tale, Moscow did not offer a legal theory. It offered denial. The Russian model treats the legal debate as irrelevant noise. There is no claimed compliance with imminence or proportionality because there is no claim at all, only the flat assertion that nothing happened and the implicit message that the target audience, future defectors, should understand otherwise. Russia matters to this discussion as the boundary case, the program that does not even pretend, and its presence at the edge clarifies what the other states are doing when they reach for legal language.

Britain’s approach is a study in distance. The United Kingdom maintains, in its domestic legal culture and its public posture, a strong presumption against extrajudicial killing. Yet Britain has conducted lethal strikes against its own nationals abroad, notably a 2015 drone strike in Syria that killed a British citizen, and Britain has supplied intelligence to allied programs whose lethal use it does not control. The British position is therefore double. It condemns the practice in principle and participates in it in fact, and it manages the contradiction through the doctrine of imminence and through the careful separation of intelligence sharing from trigger pulling.

The Indian approach, the subject the whole series circles, is the quietest of the five and in some ways the most extreme. India neither acknowledges a targeted killing program nor articulates a legal theory for one. Where Israel litigates and the United States memos, India simply denies, and the direct comparison between the two democracies shows how far apart self-conscious legality and total deniability can sit. India has no published doctrine, no court ruling, no released opinion, no white paper. The pattern of eliminations across Pakistani cities is visible, the global legal questions it raises are real, and yet the Indian state offers nothing for a lawyer to argue with. That silence is itself a legal posture, and understanding why a government would choose it is one of the things the six standards below help explain.

These five are not meant to be ranked. They are meant to be held in the same frame, because each one solves the legal problem differently and the differences expose what the problem actually is.

One way to read the five together is as a spectrum of avowal. At one end stands the fully avowed program, in which a state acknowledges that it conducts targeted killings, publishes the standards it claims to apply, and submits the practice to some form of domestic review. At the other end stands the fully deniable program, in which a state acknowledges nothing, publishes nothing, and submits to no review because there is officially nothing to review. Israel sits near the avowed end, the United States somewhat behind it, Britain in an ambiguous double position, Russia near the deniable end with the added feature of active denial, and India at the deniable end with a silence so complete it does not even bother to deny in detail. The striking thing about this spectrum is that legality, in the ordinary sense, does not track position on it. The most avowed program is not the most lawful and the most deniable is not the most lawless. What position on the spectrum tracks is exposure. The avowed state is exposed to argument, to challenge, to the discipline of having to defend itself. The deniable state is shielded from all of it. Holding that observation in mind through the six standards that follow is the key to understanding why the legal debate produces the political results it does.

It is worth being precise, at the outset, about what the standards are and where they come from, because the rest of the analysis depends on keeping them distinct. Three of the six, imminence, necessity, and self-defense, belong primarily to the law governing when a state may resort to force at all, the body of rules that radiates outward from Articles 2(4) and 51 of the United Nations Charter. Two of them, proportionality and combatant status, belong primarily to the law governing how force may be conducted once a conflict exists, the law of armed conflict. The sixth, sovereignty, is the principle that the first body of law exists to protect and the second body of law largely brackets. Targeted killing is uniquely difficult because it does not sit cleanly in either body of law. A single motorcycle-borne shooting in a foreign city implicates the resort-to-force rules, because it is force used on another state’s territory, and the conduct-of-hostilities rules, because it is a lethal strike against a particular individual, and the sovereignty principle, because it crosses a border without the host’s consent. Most acts of state violence engage one of these frameworks. Targeted killing engages all of them at once, and the standards pull against each other.

The Imminence Standard and the Stretching of Time

Begin with imminence, because imminence is where the stretching is most visible and most consequential. The right of self-defense, as nations have understood it since long before the United Nations existed, is a right to respond to a threat that is about to materialize. The classic formulation comes from a diplomatic exchange in 1837 over an incident on the Niagara River, when the American secretary of state Daniel Webster wrote that a lawful act of self-defense required a necessity that was instant and overwhelming, leaving no choice of means and no moment for deliberation. Webster was describing a steamboat about to be used by raiders. The threat in his formulation was measured in hours.

Modern targeted killing cannot survive Webster’s definition, and so the definition has been quietly rebuilt. The militant shot outside the Karachi mosque was not about to launch an attack that afternoon. He may not have been about to launch an attack that month. What he represented was a continuing capacity, a node in a network that planned operations on timelines of years. If imminence still meant what Webster meant, none of the targeted killings examined across this series would qualify as self-defense, and the legal argument would collapse at the first step. So the argument does not let imminence mean what Webster meant.

The reconstruction took two forms. The first is the idea of the continuing threat. On this view, a senior operative of an organization that has attacked before and intends to attack again presents a threat that is always, in a sense, imminent, because the planning never stops and the next operation is always in some stage of preparation. The threat is not a single event approaching on a clock. It is a standing condition. The second form is the idea of the last window of opportunity. On this view, imminence should be measured not from the moment of the threatened attack but from the moment of the last feasible chance to prevent it. If a target can be located today and cannot be located next week, then today is imminent, regardless of when the attack he is planning would have occurred, because today is when prevention is possible.

The most influential articulation of this reconstructed imminence came from Daniel Bethlehem, a former principal legal adviser to the British Foreign Office, who published a set of principles in 2012 in a leading international law journal. Bethlehem proposed that imminence should be assessed by reference to several factors, including the nature and immediacy of the threat, the probability of an attack, whether the anticipated attack is part of a concerted pattern of continuing armed activity, the likely scale of the attack, and the absence of other means of prevention. Notice what that list does. It takes a concept that once meant a clock running down and turns it into a multi-factor judgment in which the clock is only one factor among many and not the decisive one. Under the Bethlehem principles, a threat can be imminent even if no specific attack is scheduled, provided the other factors point toward action.

Defenders of the reconstruction make a serious argument, and it deserves to be stated at its strongest. The Webster standard, they say, was written for a world of state armies and visible mobilization, a world in which the approach of a threat could be watched and timed. Non-state networks do not mobilize visibly. They do not declare. A rule that forces a state to wait until an attack is hours away forces it to wait until prevention is impossible, because the planning of a terror attack is deliberately concealed until the moment of execution. To insist on strict imminence, on this argument, is not to protect anyone. It is to convert the right of self-defense into a right that can be exercised only after the bodies are counted. Bethlehem and those who think like him are not trying to abolish imminence. They are trying to make it operational against an adversary that strict imminence cannot reach.

Critics make an argument that is equally serious. Imminence, they say, was never merely a timing rule. It was a discipline. Its function was to keep self-defense tethered to genuine necessity and to prevent the right from sliding into a general license to kill anyone a government regards as dangerous. Once imminence is redefined as a standing condition or a last opportunity, the discipline is gone. Every senior figure in every militant organization is permanently imminent, because the network is always planning something. The redefinition does not adapt the rule to a new threat. It dissolves the rule. Philip Alston, the legal scholar who served as the United Nations special rapporteur on extrajudicial executions and produced an influential study of targeted killing and drones, warned precisely about this dissolution. His concern was that an elastic imminence would let states substitute a general assessment of dangerousness for the specific, evidenced, time-bound judgment that the law was supposed to require, and that once that substitution was normalized there would be no principled stopping point.

The honest reader cannot fully resolve this. Both sides are describing something true. The strict standard genuinely does fail against networks that conceal their planning. The elastic standard genuinely does erase the constraint. What the disagreement reveals is that imminence was asked to do a job it was not designed for. It was designed to regulate the timing of force between states. It was not designed to regulate the targeting of individuals inside networks, and when it is pressed into that service it either fails to permit necessary action or fails to forbid unnecessary action. There is no version of the imminence standard that does both jobs well, and that is the first of the structural gaps.

India’s silence on imminence is instructive here. Because India offers no legal theory, India never has to specify which version of imminence its operations rely on. The United States, having published, can be held to its published reasoning and criticized when a strike does not fit it. India, having published nothing, cannot be pinned to any standard and therefore cannot be shown to have violated one. The deniability that looks like a diplomatic choice is also, on inspection, a legal strategy. A state that never states a test can never be caught failing it.

It helps to see how the imminence dispute plays out across the comparison states, because each one has resolved it in a slightly different way and the differences are revealing. Israel, through its 2006 ruling and the practice around it, has largely abandoned a clock-based imminence in favor of the continuing-participation idea, holding that a member of a terror organization who repeatedly takes part in hostilities can be targeted in the intervals between specific acts. The United States, in the public portions of its legal architecture, leaned heavily on the elastic version, with one senior official describing the relevant concept of imminence as one that did not require clear evidence that a specific attack would take place in the immediate future, a formulation that critics seized on as proof that the word had been emptied of content. Britain, in defending its 2015 strike in Syria, invoked imminence explicitly and argued that the threat from the man it killed was both current and continuing, a phrasing that tried to keep the word while loosening it. Russia did not invoke imminence at all, because Russia invoked nothing. And India, characteristically, has neither affirmed nor denied any version. What the survey shows is that no two states use the word the same way, that the states which use it have all stretched it, and that the one feature common to every avowed program is that its imminence is broader than Webster’s. The strict standard has effectively no defenders among the states that actually conduct these operations. It survives only in the arguments of the critics, which is itself a piece of evidence about where the practice has settled.

There is a deeper point lurking in the imminence dispute, and it is worth drawing out because it recurs with every standard. The argument is not really about timing. It is about whether the lawfulness of killing a particular person should depend on what that person is about to do or on what that person is. A clock-based imminence ties lawfulness to conduct, to a specific impending act. A status-based or continuing-threat imminence ties lawfulness to identity, to membership and role in a dangerous organization. The law of armed conflict is comfortable targeting people for what they are, that is what combatant status means, but the law governing resort to force, the law of self-defense, was built around responses to acts. Stretching imminence is, at bottom, an attempt to import a status logic into a body of law designed for an act logic, and the friction that produces is not a drafting error that careful lawyers can smooth away. It is the friction of two different theories of when killing is permitted, forced into the same word.

Proportionality and the Arithmetic of Acceptable Death

The second standard is proportionality, and proportionality carries a confusion that has to be cleared away before the debate makes sense. The word means two different things in two different bodies of law, and the two meanings are often run together to the benefit of whoever is doing the running.

In the law governing when a state may use force at all, the law that flows from the United Nations Charter, proportionality asks whether the overall scale of a defensive response is proportionate to the threat being answered. A small provocation does not justify a large war. In the law governing how force may be used once a conflict exists, the law of armed conflict, proportionality asks something narrower and more arithmetic. It asks whether the expected harm to civilians from a particular attack is excessive in relation to the concrete and direct military advantage anticipated from that attack. The first proportionality is about the size of the campaign. The second is about the cost of the single strike.

Targeted killing lives mostly in the second proportionality, the strike-level calculation, and that calculation is where the genuine moral content of the practice sits. When a state decides to kill a militant commander, it is rarely choosing between killing him and harming no one. It is usually choosing among options that each carry some risk to bystanders. A drone strike on a compound may kill the target and also the people around him. A missile has a blast radius. Even a close-range shooting in a crowded street carries the risk of a stray round, a panicked crowd, a second target who turns out to be a relative rather than a fighter. Proportionality is the rule that is supposed to govern that risk, and the rule is genuinely hard to apply because both sides of the comparison resist measurement.

Consider the civilian-harm side first. How many bystander deaths are too many for one commander? The law gives no number, and it cannot, because the answer depends on facts that change from case to case. But the absence of a number is exploited. After a strike that kills civilians, the state that conducted it almost always describes the dead as militants, and the state on whose soil the strike occurred almost always describes them as civilians, and there is frequently no neutral party with access to the site to adjudicate. The drone campaign in Pakistan generated exactly this pattern for a decade. Casualty estimates varied not by tens but by thousands, depending on the source, because the basic question of who counted as a civilian was answered differently by different counters. One controversial American practice, reported by journalists and never fully denied, was to treat all military-age males in a strike zone as combatants unless posthumous intelligence proved otherwise. That practice did not lie about the numbers. It changed the definition that produced the numbers, and a proportionality calculation built on a contested definition of the denominator is not really a calculation at all.

Now consider the military-advantage side, which is stretched just as hard in the other direction. Proportionality compares civilian harm to concrete and direct military advantage. The word concrete is doing work. It is supposed to exclude vague, speculative, long-term benefits and limit the comparison to the specific tactical gain of the specific strike. But the advantage claimed for killing a senior militant is rarely concrete in that narrow sense. The claimed advantage is degradation of the network, disruption of planning, deterrence of others, the slow attrition of an organization’s leadership. Those are real effects, and a state is not wrong to value them. But they are not concrete and direct in the way the law’s drafters meant. They are diffuse and cumulative. When a government weighs three or four civilian deaths against the disruption of a terror network, it is placing a measurable, immediate, certain harm on one side of the scale and a diffuse, future, uncertain benefit on the other, and calling the comparison proportionality lends it a precision it does not possess.

The defenders of the practice respond that this objection proves too much. Every use of military force, they say, involves weighing certain present harm against uncertain future benefit. That is the nature of strategy. To demand that military advantage be reduced to a tactical certainty before any strike is permitted would forbid not just targeted killing but most of warfare. The proportionality rule was always a structured judgment rather than an equation, and the fact that it cannot be reduced to arithmetic does not make it meaningless. A reasonable commander, acting in good faith on the information available, can distinguish a strike whose civilian cost is plainly excessive from one whose cost is plainly acceptable, and the hard middle cases do not erase the easy ones.

That response has force, and the analysis here accepts part of it. Proportionality is not worthless. It does rule out the plainly excessive strike, the missile fired into a wedding to reach one guest. What it does not do, and cannot do, is constrain the middle, and the middle is where targeted killing lives. The Israeli Supreme Court, in its 2006 ruling, was unusually candid about this. It held that proportionality applied, that it was a real limit, and that whether a particular strike satisfied it could be determined only case by case, after the fact, on the specific evidence. That is an honest description of the standard. It is also an admission that the standard provides no guidance in advance. A rule that can be applied only retrospectively, by an investigator with full information, is not a rule that disciplines the decision to strike. It is a rule that judges the decision once the strike is history, and most targeted killings are never investigated by anyone with the access to apply it.

The motorcycle method that recurs through the shadow war is worth pausing on here, because it represents a particular answer to the proportionality problem. A drone strike on a building maximizes the risk to bystanders. Two gunmen shooting one man at close range minimizes it. Whatever else is true about the shadow war’s operational choices, the close-range method keeps the civilian-harm side of the proportionality ledger small, and a state that wanted to make a proportionality argument for such killings would have an easier time than a state defending an airstrike. India does not make that argument, because India makes no argument. But the method itself, intentionally or not, is shaped in a way that would survive proportionality scrutiny better than the alternatives, and that is one reason the comparison with the aerial campaign repays attention.

There is a further wrinkle in the proportionality analysis that the close-range method exposes, and it concerns the difference between the precision of a weapon and the precision of the intelligence that aims it. A shooting at arm’s length is precise in the mechanical sense, the round goes where it is pointed and the blast radius is effectively zero. But proportionality is not only about whether the weapon strikes the intended body. It is also about whether the intended body was the right one. A targeted killing built on faulty identification kills a precisely chosen wrong person, and the precision of the method does nothing to redeem the error. The history of every targeted killing program contains cases of mistaken identity, and the most precise methods are not immune, because the precision lives in the trigger and the error lives in the file. The Lillehammer affair, in which Israeli operatives shot a Moroccan waiter they had wrongly identified as a Palestinian organizer, is the canonical illustration, and it is examined more fully in the account of Israel’s early operations. The lesson for proportionality is that the standard’s civilian-harm side includes not only the bystanders a state knowingly accepts but also the wrong targets it unknowingly selects, and the second category is invisible at the moment of decision, because a state that knew the target was wrong would not strike. Proportionality, in other words, is applied on the basis of beliefs about identity that are themselves uncertain, and the uncertainty does not appear anywhere in the calculation.

The proportionality standard also interacts with the question of who conducts the assessment, and here the avowal spectrum returns. For an avowed program, the proportionality judgment is at least potentially reviewable. Israel’s framework includes the requirement of an after-the-fact examination. The American framework, once its memos became public, became subject to litigation and oversight in which proportionality figured. The judgment is still made by the striking state, but it is made knowing that someone may later look. For a fully deniable program, no one looks, because there is officially no strike to assess. The proportionality calculation, in that situation, is made entirely inside the head of the official ordering the operation, with no prospect of external review and no record that could ever be examined. A standard that depends on good-faith judgment is only as good as the incentives surrounding the judge, and a deniable program removes every incentive that external scrutiny would supply. This is not an argument that deniable programs reach worse proportionality outcomes in fact. It is an argument that the standard, as a discipline, simply does not operate on them, because the discipline of proportionality was always the discipline of a judgment that might be checked.

Necessity, Last Resort, and the Capture Alternative

The third standard is necessity, and necessity in this context narrows quickly to a single hard question. If a state could capture a militant and try him, is it permitted to kill him instead? The principle of necessity, sometimes phrased as a requirement of last resort, says that lethal force may be used only when no less harmful option is available. Applied to targeted killing, that principle would seem to require a government to capture rather than kill whenever capture is feasible. The entire architecture of the practice depends on how the word feasible is interpreted, and feasible turns out to be one of the most elastic words in the whole debate.

Start with the cases where capture genuinely is not feasible, because they are real and they matter. A militant living openly in a city controlled by a state that protects him cannot be arrested by the state that wants him. There is no warrant that the host country will execute, no extradition treaty that will be honored, no police force that will cooperate. The decades of failed efforts to bring the organizers of the Mumbai attacks to trial illustrate the point with painful clarity. The legal processes existed. The charge sheets were filed. The international designations were issued. And the men named in them continued to live, in some cases publicly, because the state on whose soil they lived would not move against them. In that situation, a demand that India capture rather than kill is not a demand for a less harmful option. It is a demand that India do nothing, because capture is not on the menu. The infeasibility of capture, in cases like these, is not an excuse a government invents. It is a fact the host state has created.

The American framework built this directly into its three-part test. One of the three conditions for a lethal strike against a citizen senior operative was that capture be infeasible. The condition sounds like a real constraint, and in some applications it is. But the difficulty appeared the moment anyone asked who decides feasibility and against what standard. Feasible can mean physically possible. It can mean possible without unacceptable risk to the soldiers conducting the capture. It can mean possible without a diplomatic rupture with the host state. It can mean possible within a politically acceptable cost. Each of those readings is defensible, and each produces a different answer. A capture operation deep inside a hostile country is almost always physically possible in the narrow sense, the raid that killed Osama bin Laden proved that a capture-or-kill operation could reach into a Pakistani garrison town. But the same operation is almost never feasible in the broader senses, because it risks the lives of the assault team, risks a firefight with the host state’s forces, and risks a diplomatic crisis. So a government that wants to kill rather than capture does not have to lie about feasibility. It simply has to choose the broad reading, under which capture is rarely feasible, over the narrow reading, under which it often is.

Critics of the practice make a pointed argument here, and it is worth taking seriously rather than waving away. The necessity requirement, they say, exists precisely to prevent killing from becoming the default. If feasibility is read so broadly that any inconvenience or risk renders capture infeasible, then the requirement evaporates and lethal force becomes the first resort dressed in the language of the last. Worse, the broad reading creates a perverse incentive. A state that prefers killing to the complications of detention, the legal scrutiny of a trial, the political awkwardness of a prisoner who can talk, has every reason to declare capture infeasible, because the declaration is unfalsifiable. No one can prove that a capture the state did not attempt would have succeeded. The necessity standard, on this critique, does not constrain the choice between capture and killing. It launders it.

The defenders answer that the critics are demanding a fantasy. Capture, they say, is not a neutral, costless alternative that a humane state would always prefer. Capture means a raid, and a raid means soldiers in a foreign country, a firefight, the risk of dead troops and dead bystanders and a captured commando paraded on television. Capture means a detainee, and a detainee means the questions that detention has generated for two decades, where to hold him, under what law, with what trial, with what outcome if the evidence is intelligence that cannot be shown in court. The American detention experiment and its legal aftermath, the defenders argue, showed that capture is not the clean option the critics imagine. Sometimes a clean kill is genuinely less harmful, all things considered, than a messy capture, and a necessity standard that ignores the real costs of capture is not a serious standard.

Both arguments are partly right, and the place where they meet is uncomfortable. The critics are right that the broad reading of feasibility hollows out the necessity requirement and creates an incentive to kill. The defenders are right that capture is genuinely costly and sometimes genuinely worse. What neither side can produce is a definition of feasible that is both administrable and constraining. A narrow definition, capture is feasible whenever it is physically possible, is constraining but unrealistic, because it would forbid killing in cases where capture would cost many lives. A broad definition, capture is feasible only when it is safe, cheap, and diplomatically painless, is realistic but empty, because it permits killing in nearly every case. The standard cannot be both, and so the necessity requirement, like the imminence requirement, turns out to be a genuine principle wrapped around a hole.

For India the necessity question has a particular shape. Of all the states compared here, India faces the clearest case of genuine capture infeasibility, because the militants it is accused of targeting live in a country that has spent decades refusing to surrender them. If any state could honestly say that capture is not an option, it is India confronting a wanted man in Karachi or Lahore. Yet India says nothing, and the silence costs it something. By declining to make even the capture-infeasibility argument, the strongest legal argument available to it, India forgoes the one piece of legal high ground it could plausibly hold. That choice only makes sense if the value of total deniability outweighs the value of the best available defense, which is a revealing trade and one the final sections return to.

The necessity standard also carries a problem that is easy to miss, because it hides inside the comparison the standard requires. Necessity asks whether a less harmful option exists, and the less harmful option is almost always assumed to be capture followed by trial. But capture followed by trial is only less harmful if the trial is real, and a trial is only real if the evidence can be produced in a courtroom. Much of the intelligence that identifies a militant leader as a legitimate target is exactly the kind of intelligence that cannot be produced in open court, because producing it would expose the sources and methods that generated it. A state may possess complete certainty that a man organized a massacre and still be unable to prove it to a judge without burning the human source who reported it or the intercept capability that recorded it. In that situation the capture-and-try alternative is not genuinely available even when the physical capture is feasible, because the second half of it, the trial, would either fail for lack of admissible evidence or succeed only at an intelligence cost the state will not pay. The necessity standard, by treating capture and trial as a single clean alternative, conceals the fact that the two halves can come apart, and a serious application of the standard would have to ask not merely whether a man can be seized but whether, once seized, he could be lawfully convicted on evidence the state is willing to disclose. That question is rarely asked, and when it is asked honestly the answer often narrows the genuine alternatives more than the defenders of capture would like to admit.

There is one more dimension to necessity that the comparison states illuminate, and it concerns the relationship between necessity and tempo. A method that kills one target at a time, at close range, is inherently low-tempo. It cannot be scaled the way an aerial campaign can. This has a curious effect on the necessity analysis. A low-tempo method forces a state to be selective, because it cannot strike everyone, and selectivity pushes the state toward targeting only the figures whose elimination genuinely matters, the planners and organizers rather than the foot soldiers. A high-tempo method removes that discipline, because when striking is cheap and frequent the pressure to reserve it for the genuinely necessary case relaxes. The drone campaign’s critics made exactly this point, arguing that the very efficiency of the method eroded the necessity filter, that strikes which would never have been ordered if each one required a difficult, costly operation were ordered routinely once the method made them easy. The close-range method, whatever its other features, does not have this problem, because its difficulty is its own discipline. This is not a defense of the method. It is an observation that necessity, as a practical constraint, depends partly on how expensive the lethal option is, and that the cheapest methods are the ones on which the necessity standard exerts the least real pressure.

Sovereignty, Self-Defense, and the Unwilling or Unable Doctrine

The fourth and fifth standards, sovereignty and self-defense, cannot really be separated, because the whole modern debate over targeted killing on foreign soil is a debate about how to reconcile them. They sit in tension at the center of the United Nations Charter, and the doctrine that has grown up to manage the tension, the unwilling or unable doctrine, is the single most important and most contested idea in this entire field.

The tension is built into the Charter’s text. Article 2(4) commits every member state to refrain from the threat or use of force against the territorial integrity or political independence of any other state. It is the foundation stone of the post-1945 order, the rule that makes the borders mean something. Article 51 preserves, against that prohibition, the inherent right of individual or collective self-defense if an armed attack occurs against a member state. Read together, the two articles describe a clear world. You may not use force on another state’s territory, except that you may defend yourself if that state attacks you. For most of the Charter’s history, the exception was understood to require an armed attack by a state. The self-defense in Article 51 was self-defense against a country.

What targeted killing forces into the open is a different problem, the problem of the non-state attacker operating from a state that is not itself the attacker. A militant network launches an armed attack from inside Country B against Country A. Country B did not order the attack and is not, in the legal sense, responsible for it. But Country B also does not stop the network, either because it cannot, its writ does not run in the relevant territory, or because it will not, the network serves the host’s purposes or enjoys the protection of its security establishment. Country A now faces a continuing threat from a group it cannot reach without entering Country B’s territory, and entering Country B’s territory with force is exactly what Article 2(4) forbids. The Charter, read strictly, leaves Country A with no lawful military option against the network, because the network is not a state and Country B, the only state in the picture, has not committed an armed attack.

The unwilling or unable doctrine is the answer that powerful states have constructed to escape that trap. The doctrine holds that when a state is unwilling or unable to suppress a threat emanating from its territory, the victim state may act in self-defense against the non-state group on that territory, and the host state’s sovereignty does not bar the action. The host has, in effect, forfeited the protection of Article 2(4) with respect to the specific threat it has failed to address. The host is not being attacked and is not the target. The network is the target, and the host’s sovereignty yields only so far as necessary to reach it.

This doctrine has a real pedigree and a serious logic. Its defenders argue that it is the only reading of the Charter that does not produce an absurd result. The alternative, strict reading would mean that a state can immunize a terror network simply by being too weak or too complicit to act against it, and that the network’s victims have no lawful remedy but to absorb attack after attack. That cannot be what the law intends, the defenders say, because a legal order that rewards a host state’s failure by stripping the victim of any defense is not a legal order anyone would design. Daniel Bethlehem’s principles, mentioned earlier in connection with imminence, also gave the unwilling or unable doctrine its most careful modern statement, and the structure of the argument is that Article 51’s right of self-defense, being inherent, must be capable of being exercised against the actual source of the armed attack, which in the modern world is frequently a non-state group rather than a government.

The critics’ response is layered, and each layer matters. The first objection is textual. Article 51 speaks of an armed attack, and for most of the Charter’s life the International Court of Justice and the weight of scholarly opinion read armed attack to mean an attack attributable to a state. On that reading, an attack by a non-state group that cannot be attributed to the host government simply does not trigger Article 51 at all, and the unwilling or unable doctrine is not an interpretation of the Charter but a departure from it. The second objection is about consent and standards. Who decides that a host state is unwilling or unable? The acting state decides, by itself, about itself. There is no neutral arbiter, no agreed threshold, no procedure. A state that wants to strike inside another’s territory simply announces that the host is unwilling or unable, and the announcement is both the judgment and the justification. The third objection is about precedent and power. The unwilling or unable doctrine, the critics note, is invoked almost exclusively by militarily powerful states against weaker ones. It is never the weak state that declares the strong one unable. The doctrine, on this view, is not a neutral principle but a legal dress for the old reality that strong states use force on weak states’ territory and need a vocabulary that makes it sound lawful.

There is also a quieter objection that deserves its own sentence, because it is the one that should trouble even a sympathetic reader. The doctrine has no natural limit. Once a state may strike a non-state group on the territory of any host it deems unwilling or unable, the geographic scope of armed force expands to wherever the group has a presence, and the temporal scope expands to however long the group exists. The conflict has no front line and no end date. The unwilling or unable doctrine, whatever its logic in the individual case, dissolves the boundaries that the law of armed conflict was built around, and a body of law without boundaries struggles to constrain anything.

Where the analysis lands is, again, in the uncomfortable middle. The strict reading of the Charter genuinely does produce the absurd result its critics deny and its defenders emphasize, a victim state with no remedy against a network its host shelters. The unwilling or unable doctrine genuinely does cure that absurdity, and it does so with a logic that is not merely a fig leaf. But the doctrine also genuinely does what its critics fear. It hands the decisive judgment to the acting state, supplies no standard, sets no limit, and operates in practice as a tool of the strong. Both things are true at once. The doctrine is the least bad available answer to a real problem, and it is also an open door that the law cannot close once it has been walked through.

India’s relationship to this doctrine is the most revealing of all, and it explains a great deal about the shadow war’s design. India could invoke the unwilling or unable doctrine. The factual predicate is arguably the strongest of any state discussed here. India can point to a decades-long record, examined in Pakistan’s track record as a sheltering state and elsewhere in this series, of attacks launched from a neighbor’s soil by groups the neighbor’s security establishment has protected rather than suppressed. A government building a public legal case for cross-border action would build it on exactly this foundation. India does not build it. India invokes no doctrine, claims no self-defense, names no unwilling or unable host. It denies that there is anything to justify.

That choice looks strange until the incentives are laid out, and then it looks rational. Invoking the unwilling or unable doctrine would require India to admit the operations, and admitting the operations would convert a deniable pattern into an avowed policy. An avowed policy invites a legal response, a diplomatic response, a demand for evidence, a forum in which the host state can litigate the unwilling or unable judgment that India would have to assert. Denial avoids all of it. By saying nothing, India keeps the operations in a zone where they can be neither legally defended nor legally condemned, because there is no official act to defend or condemn. The unwilling or unable doctrine is a tool for states that want to act openly and need legal cover. India has decided it would rather act silently and need no cover at all. That is the deepest sense in which the legal ambiguity is a feature: it is the thing that makes a no-doctrine policy survivable.

The sovereignty side of the equation deserves a closer look, because it is often treated as the simple half of the pairing and it is not. Sovereignty is not only a shield that a host state holds up. It is also a claim about who is responsible for what happens on a given territory. When the unwilling or unable doctrine is invoked, it does something subtle to that claim. It says, in effect, that a host state which fails to police its own territory cannot then assert the full protection of sovereignty over the part of the territory it has failed to police. The doctrine ties the shield to the duty. Hold up the shield, the doctrine says, only if you have performed the duty the shield presupposes. There is a real moral logic in that, and it is the logic that makes the doctrine attractive even to people uneasy about its consequences. A state that knowingly shelters a terror network, that lets its security establishment protect the network’s leadership, has arguably already breached the sovereignty bargain, and there is something hollow in letting it then invoke sovereignty to bar the network’s victims from acting. The history examined across this series, the long record of attacks launched from a neighbor’s soil by groups that neighbor’s institutions protected, is precisely the kind of record the unwilling or unable doctrine was built to address.

But the same logic, followed one step further, exposes the doctrine’s danger. If sovereignty can be forfeited by a failure of policing, then the question of whether sovereignty has been forfeited becomes a question of fact, and that question of fact is decided, in the current system, by the very state that wants to cross the border. There is no tribunal that rules on whether a host is unwilling or unable. There is no agreed evidentiary threshold. There is only the acting state’s own assessment, announced to the world as both finding and justification. A host state genuinely trying and genuinely failing, a weak state whose writ does not reach a remote province, is treated the same as a host state cynically protecting a network it finds useful, because the doctrine as currently used does not reliably distinguish unable from unwilling, and the acting state has every incentive not to look too hard at the difference. The doctrine that began as a way to tie the shield to the duty ends, in practice, as a way for the militarily strong to decide unilaterally that the militarily weak have forfeited protection. Both the moral logic and the danger are real, and the honest reader has to hold them together rather than choosing the one that is more comfortable.

It is also worth noting how rarely the unwilling or unable doctrine has been tested in any forum that could rule on it. The International Court of Justice has had occasion to touch the question of self-defense against non-state actors and has, on the whole, given answers that sit awkwardly with the doctrine, suggesting in more than one decision that an armed attack triggering Article 51 must have a sufficient connection to a state. The doctrine’s defenders argue that state practice has moved on, that the court’s pronouncements predate the era of transnational networks, and that what states actually do, and what other states actually accept, now constitutes a customary acceptance of the doctrine. The doctrine’s critics argue that practice by a handful of powerful states, acquiesced in by their allies and protested by everyone else, is not the kind of general practice that makes customary law. This dispute, about whether the doctrine is already law or merely a powerful claim about what the law should be, has no neutral arbiter either, and so it joins the imminence dispute, the proportionality dispute, the necessity dispute, and the combatant status dispute as one more place where the law on targeted killing does not resolve but merely relocates the disagreement.

Combatant Status and the Problem of the Civilian Who Fights

The sixth standard is combatant status, and it raises a question that sounds technical and turns out to be the moral hinge of the whole subject. International humanitarian law divides people into categories. Combatants may be targeted at any time, because being a lawful object of attack is part of what it means to be a combatant. Civilians may not be targeted at all, because protection from attack is the core of civilian status. The division is clean, and it works tolerably well for the kind of conflict it was designed for, two uniformed armies facing each other across a line. It works badly for a militant network, because the people in a militant network are not, in the legal sense, combatants. They wear no uniform, carry arms openly only when they choose to, belong to no state’s armed forces. By the strict letter of the categories, a member of a terror organization is a civilian.

But a civilian who plans and directs attacks on other civilians cannot simply be left in the protected category, because that would mean a terror commander enjoys, as a matter of law, an immunity from attack that a uniformed soldier does not. The law’s answer to this is the concept of direct participation in hostilities. A civilian who takes a direct part in hostilities loses, for the duration of that participation, the protection that civilian status would otherwise provide. The civilian becomes targetable, not by being reclassified as a combatant, but by forfeiting protection through conduct. The Israeli Supreme Court’s 2006 ruling rested heavily on this concept, holding that members of terror organizations were civilians who could be targeted only insofar as and only for such time as they were directly participating in hostilities.

The phrase for such time is where the difficulty concentrates. A civilian who fires a weapon is directly participating while he fires and loses protection during that act. But what about the commander who plans the operation and never touches a weapon? What about the financier, the recruiter, the bomb maker, the man who provides safe houses? Are they directly participating, and if so, for how long? If direct participation is read narrowly, limited to the act of fighting itself, then most of a network’s leadership is never targetable, because leaders do not fight. If it is read broadly, extended to anyone who contributes to the organization’s violent capacity, then the category of the targetable swells until it includes people whose connection to any specific act of violence is remote.

This was the question the International Committee of the Red Cross tried to answer in a major study published in 2009, an interpretive guidance on the notion of direct participation in hostilities. The study, whose principal author was the legal scholar Nils Melzer, who would later write the leading academic book on targeted killing in international law, attempted to draw the line with three criteria. An act amounts to direct participation, the guidance proposed, when it is likely to harm the adversary or protected persons, when there is a direct causal link between the act and that harm, and when the act is specifically designed to support one party against another. The guidance also introduced a more controversial idea, that a person who assumes a continuous combat function within an organized armed group could be targeted on the same basis as a member of a state’s armed forces, rather than only during specific acts.

The guidance did not settle the debate. It sharpened it. The continuous combat function idea was attacked from one direction by those who said it smuggled combatant status back in through a side door, allowing the targeting of people on the basis of their role rather than their conduct, which is exactly what the civilian protection rule was supposed to forbid. It was attacked from the other direction by military lawyers who said the three criteria for direct participation were too narrow, that they would protect bomb makers and financiers and planners whose contribution to violence was obvious to everyone except the law. Melzer’s guidance was a serious, careful attempt to hold a line in an impossible place, and the fact that it was attacked from both sides at once is the clearest evidence that the place is genuinely impossible.

What hangs on this is not academic. The whole legitimacy of targeting a militant leader depends on the answer. If the leader is a civilian not directly participating, killing him is the killing of a protected person and the legal defense fails entirely. If he is a civilian who has assumed a continuous combat function, or who is read as continuously and directly participating because his planning never stops, then killing him is lawful and the defense holds. The same man, the same facts, two legal characterizations, and the entire question of whether a targeted killing is a lawful act of war or an extrajudicial execution rides on which characterization a given lawyer adopts. There is no neutral fact that decides it. The categories were built for a different kind of war, and when they are applied to a network they do not so much answer the question as relocate it.

The shadow war’s targets sit squarely in this gray zone. A senior organizer of a militant group is not firing a weapon when the motorcycle pulls alongside him. He is, on a narrow reading, a civilian going about his day. He is, on a broad reading, a man performing a continuous combat function for an organization whose function is mass violence. The killing is lawful or unlawful depending entirely on which reading governs, and because India publishes no legal position, India never has to say. Once again the pattern repeats. The state that articulates a doctrine, as Israel did through its court, can be measured against it and found wanting in particular cases. The state that articulates nothing floats free of the whole framework, neither defended by it nor condemned by it.

The combatant status problem has a temporal dimension that is worth isolating, because it is where the deepest unfairness in the whole structure becomes visible. Consider a militant leader’s life across a single week. On Monday he meets a financier. On Tuesday he reviews a plan. On Wednesday he prays, eats, sees his children. On Thursday he instructs a cell. On Friday he is shopping in a market. Under a strict, act-focused reading of direct participation, he is targetable on Monday, Tuesday, and Thursday, when his conduct is plausibly participation, and protected on Wednesday and Friday, when it is not. That reading produces an almost absurd operational rule, a man who may be killed on certain days of the week and not others, with the lawfulness of a strike turning on the calendar. The continuous combat function idea was developed precisely to escape that absurdity, by holding that a person whose function within the organization is continuous can be targeted continuously, including on the Wednesday and the Friday. But escaping the absurdity comes at a price. Once a person can be targeted on the day he is doing nothing violent, lawfulness has detached entirely from conduct and attached to status, and the protection that civilian status was supposed to provide has been replaced by a role-based vulnerability that looks exactly like combatant status without the name. The law cannot have it both ways. Either lawfulness tracks conduct, and produces the calendar absurdity, or it tracks status, and erases the civilian-combatant distinction for everyone the state designates. The continuous combat function concept is the law’s attempt to occupy a middle that, on inspection, is not actually there.

There is also the question of how a person enters the targetable category and how, if ever, a person leaves it. A uniformed soldier becomes a combatant by a formal act, enlistment, and ceases to be one by another formal act, discharge or surrender or capture. A militant designated as continuously participating entered the category by an intelligence assessment and can leave it, in principle, by ceasing to participate, but who decides that he has ceased, and on what evidence? A man may have stepped back from operations, may be old, may be sidelined, may be in a dispute with his organization, and none of that is visible to the state holding the file. The intelligence that placed him in the category is rarely updated to take him out of it, because the incentives all run one way, a target removed from the list is a target who might later turn out to have been active after all, and no analyst is rewarded for that. The result is that the targetable category, once entered, tends to be permanent in practice even where the law would say it should be revocable. This is not a flaw the law acknowledges. It is a flaw in how the law’s categories interact with the institutions that apply them, and it means that the combatant status standard, like the other five, promises a precision in distinguishing the targetable from the protected that the machinery underneath it cannot actually deliver.

Where the Legal Frameworks Break Down

Step back from the six standards and a single shape becomes visible. Imminence was built to time the use of force between states and breaks when asked to time the targeting of individuals inside networks. Proportionality was built to compare measurable harms and breaks when one side of the comparison is the diffuse, future benefit of degrading an organization. Necessity was built to require the least harmful option and breaks because feasibility, the word that controls it, can be read narrowly into unrealism or broadly into emptiness. Sovereignty and self-defense were built around the armed attack of a state and break when the attacker is a network and the host is merely unwilling or unable. Combatant status was built for armies and breaks when applied to people who are civilians by the letter of the law and fighters by every other measure. Each standard is a real principle. Each is also a real principle wrapped around a hole, and the holes are not random. They appear at exactly the same place in every standard, the place where a body of law designed for war between states is asked to govern lethal action against individuals inside non-state networks.

This is the central claim of the analysis, and it should be stated plainly. The indeterminacy of the law on targeted killing is structural, not accidental. It is not the case that the law has a clear answer that lawyers have failed to find, or that a future treaty could supply the answer if states found the political will. The law is indeterminate because it was constructed for one kind of problem and is being used on another, and the mismatch produces gaps at every load-bearing point. A clarification that closed the gaps would not be a clarification at all. It would be new law, and new law on this subject would require states to agree on the one thing they have shown for fifty years they will not agree on, namely whether killing individuals on foreign soil is something sovereign states are entitled to do.

And here the argument turns, because the natural reaction to structural indeterminacy is to call it a defect and demand repair. That reaction is worth resisting. Consider what a genuine resolution would look like, in either direction. Suppose international law were clarified to permit targeted killing under defined conditions, a clear imminence threshold, a fixed proportionality formula, an agreed feasibility standard, a procedure for the unwilling or unable judgment. The permission would immediately become a template. Every state with a grievance and a capable intelligence service would have a checklist for lawful assassination, and the practice that is now controversial and constrained by reputational cost would become routine and legitimate. The clarity that disciplined the careful state would also license the reckless one, because a clear rule is a clear rule for everyone who can read it.

Now suppose the opposite, that international law were clarified to prohibit targeted killing absolutely, no exceptions, no unwilling or unable, no continuous combat function. The prohibition would not stop the killing. It would only strip away the legal vocabulary that currently surrounds it. States facing a real and unreachable threat would not absorb attack after attack out of respect for a rule. They would act, and they would act in the Russian mode, without justification, because no justification would be available. The absolute prohibition would not produce a world with fewer targeted killings. It would produce a world in which every targeted killing was, by definition, a flat violation, and in which the careful state that wanted to act within limits would have no limits to act within, because the law would offer none short of total abstention.

The current ambiguity, on this argument, does something that neither clear answer does. It permits the careful state to act while keeping the action contested. A government conducting targeted killings today operates in a zone where its conduct can be defended but not vindicated, and criticized but not definitively condemned. That zone imposes a real cost. It means every operation carries reputational risk, every strike can be challenged, every program must be conducted with at least one eye on how it will be argued about. The cost is not nothing. It is the reason states invest in legal architecture, why Israel litigates and the United States memos and even Russia bothers to deny. The ambiguity keeps targeted killing expensive enough to remain exceptional, while permitting it in the cases where a state judges the cost worth paying. A clear permission would make it cheap. A clear prohibition would make it lawless. The unclear middle keeps it costly and contested, and costly and contested may be the best a fractured international system can do with a practice that powerful states will not surrender and weaker states cannot stop.

This is not a comfortable conclusion and it is not offered as a comfortable one. It does not say targeted killing is right. It says the law’s refusal to decide is doing work, and that the work is not obviously worse than the alternatives. A reader who finds that troubling has understood the argument correctly. The discomfort is the point, because the subject does not admit of a resolution that is both honest and comfortable.

There is one place where the ambiguity’s defenders should concede ground, and the analysis concedes it here. The argument that ambiguity keeps the practice costly assumes that the cost is actually imposed, that there is scrutiny, criticism, and reputational consequence. For an avowed program, that assumption holds. Israel’s program is debated in its own courts and parliament and press. The American program, once its memos were public, became the subject of litigation and oversight. But for a fully deniable program, the assumption weakens. A state that never admits the operations faces no domestic legal challenge to them, no parliamentary inquiry into them, no journalist with standing to demand the legal theory, because officially there is nothing to inquire into. The ambiguity that keeps an avowed program costly may keep a deniable program almost free. And that observation leads directly to the place where India sits, and to what the whole comparison finally teaches.

Before turning to that, one objection to the entire structural-indeterminacy argument should be stated and answered, because it is the strongest objection available. A critic might say that the argument proves too much, that by the same reasoning every legal standard everywhere is indeterminate, since every standard requires interpretation and every interpretation can be contested. Reasonableness, due care, proportionate force, all of law is full of standards that resist mechanical application, and yet law functions. Why single out the law of targeted killing as specially or structurally indeterminate? The answer is that there is a difference between a standard that is vague and a standard that is mismatched. A vague standard, reasonableness, has a settled purpose and a community of practice that, over many cases, converges on rough agreement about what it requires. The standard is fuzzy at the edges but stable at the core. The standards governing targeted killing are not merely fuzzy at the edges. They are being applied to a kind of problem they were not built for, and the mismatch means there is no stable core for a community of practice to converge on. Imminence applied to inter-state force has a core. Imminence applied to network targeting does not, because the two readings, act-based and status-based, do not converge over time, they diverge, each pulling toward a different theory of permitted killing. That is what structural means here. The indeterminacy is not the ordinary open texture of legal language. It is the specific instability that results from forcing one body of law to do another body of law’s job, and no amount of accumulated practice will resolve it, because practice cannot settle a question the underlying framework was never designed to pose.

What the Debate Teaches

The five national approaches set out at the start can now be read as five different positions along a single axis, and the axis is not legality. It is acknowledgment. Every state on it conducts or has conducted lethal operations against individuals abroad. What separates them is how much they say.

Israel says the most. It has a court ruling, a public legal standard, a documented history, an internal debate. Israel’s program is the most constrained of the five precisely because it is the most avowed, and the longer history of Israel’s doctrine shows the constraint developing over decades. The United States says less than Israel but far more than the others, having produced a detailed legal architecture that became public, however reluctantly, and however incomplete. Britain says less again, condemning the practice in principle while participating through intelligence sharing and the occasional avowed strike, managing the gap with the language of imminence. Russia says nothing but denies, which is a kind of speech, an assertion that there is nothing to discuss. And India says nothing and does not even deny in detail, maintaining a silence so complete that there is no official act for the law to grip.

The instinctive ranking of these positions puts Israel at the top, as the most lawful, and India at the bottom, as the most lawless. The analysis resists that ranking, not because it is wrong about the law but because it misunderstands what the silence is doing. India’s silence is not an admission of lawlessness. It is a calculated refusal to enter the legal arena at all, and the refusal is rational given everything the six standards have shown. The law of targeted killing is structurally indeterminate. Any legal theory India advanced, on imminence, on the unwilling or unable doctrine, on continuous combat function, would be contestable, because every such theory is contestable. Advancing one would convert a deniable pattern into an avowed policy and hand adversaries a stated position to attack. Advancing none keeps India in the only space the indeterminate law genuinely protects, the space of the act that is never officially performed.

This is the accountability gap that the comparison exposes, and it cuts in a direction that should unsettle defenders and critics alike. The avowed program, Israel’s, is the most legally exposed and therefore the most constrained, scrutinized by its own institutions and answerable when a strike goes wrong. The deniable program, India’s, is the least legally exposed and therefore the least constrained, because there is no admitted policy for any institution to scrutinize. The structure of international law, by leaving targeted killing in the indeterminate middle, rewards the state that says least. It imposes its costs, reputational risk, legal challenge, the discipline of having to argue, on the government honest enough to acknowledge what it does, and it imposes almost none of those costs on the government that simply refuses to acknowledge. The honest state pays for its honesty. The silent state is subsidized for its silence.

That is the teaching, and it is a bleak one. The legal debate over targeted killing is not a debate between a permission and a prohibition. It is a debate among states that have all made the same choice about whether to kill and have made different choices about whether to admit it, conducted under a body of law that, by failing to decide the underlying question, has accidentally arranged its incentives to favor concealment over candor. India’s shadow war is not an aberration from this system. It is the system’s most logical product, the strategy a rational state adopts once it understands that the law will neither bless nil action nor forbid it, and that the safest place to stand in an indeterminate legal landscape is the place where you have said nothing it can hold against you.

The question that opened this article, whether the killing outside the Karachi mosque was lawful, therefore has an answer, though not the answer either camp wants. The answer is that the law does not say, that its silence is structural and probably permanent, and that the silence is itself the most important fact about the practice. The frameworks examined here, imminence and proportionality and necessity and sovereignty and self-defense and combatant status, are not failed attempts at clarity. They are the working machinery of a deliberate ambiguity, and the ambiguity is what has allowed targeted killing to persist for half a century as something the powerful do, the law tolerates, and no one is ever quite able to call by a single name.

It remains to say what follows from this for anyone trying to think clearly about the shadow war or about any state’s targeted killing program. The first thing that follows is a discipline of language. Because the law does not decide, the words used to describe these killings, assassination, elimination, neutralization, counter-terror strike, extrajudicial execution, are not neutral reports of a legal fact. They are arguments. Each word smuggles in a conclusion about a question the law has left open, and a careful reader should treat the choice of word as evidence of the speaker’s position rather than as a description of the event. The second thing that follows is a caution about certainty. Anyone who tells you, with confidence, that a particular targeted killing was definitely lawful or definitely unlawful has either not understood the six standards or has decided to suppress what understanding them reveals. The honest position is the uncomfortable one, that the killing can be argued either way, that the arguments on both sides are real, and that no neutral authority exists to choose between them. The third thing that follows is the most important, and it is a point about power. The indeterminacy is not neutral in its effects. It systematically advantages the states with the capability to act and the diplomatic weight to absorb criticism, and it systematically advantages, among those states, the ones that say least. A reader who absorbs only the conclusion that the law is unclear has absorbed half the lesson. The other half is that unclear law is not empty law. It is law whose gaps are occupied, and they are occupied by whoever is strong enough and quiet enough to occupy them.

Frequently Asked Questions

Is targeted killing ever legal under international law?

There is no single answer, and that absence is the central fact. International law contains no provision that clearly permits targeted killing and none that clearly forbids it. Whether a particular killing is lawful depends on how a series of contested standards is interpreted, imminence, proportionality, necessity, sovereignty, self-defense, and combatant status, and reasonable lawyers reach opposite conclusions on the same facts. A state can construct a legal defense for almost any individual strike, and a critic can construct a legal objection to the same strike. The law’s refusal to decide is not a temporary gap awaiting a fix. It is a structural condition of a body of rules built for war between states and applied to lethal action against individuals inside non-state networks.

What is the unwilling or unable doctrine?

It is the legal theory that allows a state to use force against a non-state group on another state’s territory when the host government is either unwilling to suppress the group or unable to do so. The idea is that a host which fails to address a threat emanating from its soil cannot use its sovereignty to shield that threat from the victim state’s self-defense. The doctrine resolves a real problem, a victim state otherwise has no lawful military remedy against a network its host protects, but it also hands the decisive judgment to the acting state, supplies no neutral standard for deciding when a host is unwilling or unable, and is invoked almost exclusively by militarily powerful states against weaker ones.

How do different countries legally justify targeted killings?

They justify them very differently, and the differences are mostly about how much each state says rather than what each state does. Israel produced a Supreme Court ruling in 2006 that assessed targeted killing under the law of armed conflict and required case-by-case justification. The United States built a detailed set of legal memos inside its Department of Justice, including a three-part test of an informed high-level official, infeasible capture, and law-of-war compliance. Britain condemns the practice in principle while participating through intelligence sharing and occasional avowed strikes. Russia offers no legal justification and relies on denial. India offers neither a justification nor a detailed denial, maintaining a silence so complete that there is no official position to evaluate.

Does the current legal ambiguity serve a purpose?

It can be argued that it does, though the argument is uncomfortable. A clear legal permission would turn targeted killing into a routine, templated practice available to every state with a capable intelligence service. A clear legal prohibition would not stop the killing but would strip away the legal vocabulary that currently surrounds it, pushing states toward the unjustified, denial-only model. The ambiguous middle keeps the practice costly and contested, imposing reputational risk and the discipline of having to argue, while still permitting action in cases a state judges worth the cost. The ambiguity’s defenders say this is the least bad outcome. Its critics say the costs fall unevenly and reward concealment.

Should international law be clarified on targeted killing?

The case against clarification is stronger than it first appears. Clarification would require states to agree on whether killing individuals on foreign soil is a legitimate sovereign act, and they have shown for fifty years that they will not agree. A clarification that permitted the practice would license reckless states along with careful ones. A clarification that prohibited it absolutely would not end the killing but would make every instance a flat violation, removing any framework of limits for states determined to act. The structural indeterminacy is frustrating, but every proposed cure has serious side effects, and no clarification can deliver discipline without also delivering either a license or a fiction.

What is the UN position on extrajudicial killing?

United Nations human rights bodies, and successive special rapporteurs on extrajudicial, summary, or arbitrary executions, have consistently expressed concern that targeted killing programs risk violating the right to life and eroding the legal limits on the use of force. Philip Alston, who held the special rapporteur role and produced an influential study of targeted killing and drones, warned that an elastic reading of imminence and self-defense could let states substitute a general assessment of dangerousness for the specific, evidenced judgment the law requires. The UN position is best described as sustained normative pressure rather than a binding determination, because the relevant bodies can register objection but cannot resolve the underlying legal indeterminacy.

Can a targeted killing ever satisfy proportionality requirements?

In principle yes, and in the easy cases the standard does real work. A strike whose civilian cost is plainly excessive, a missile fired into a crowd to reach one man, fails proportionality clearly. The difficulty is the middle range, where the standard provides almost no guidance. Proportionality compares expected civilian harm against concrete and direct military advantage, but the advantage claimed for killing a senior militant, degradation of a network, disruption of planning, deterrence, is diffuse and future rather than concrete and direct. Comparing a certain present harm against a diffuse future benefit is a structured judgment rather than a calculation, and a close-range method that minimizes bystander risk satisfies it more comfortably than an airstrike does.

Why does India operate without a stated legal framework?

Because, given the structure of the law, silence is the rational strategy. Any legal theory India advanced would be contestable, since every theory in this field is contestable, and advancing one would convert a deniable pattern into an avowed policy that adversaries could attack in legal and diplomatic forums. Total deniability keeps the operations in a zone where they can be neither legally defended nor legally condemned, because there is no official act to defend or condemn. The cost is that India forgoes the strongest argument available to it, capture infeasibility, but the calculation evidently treats the value of deniability as higher than the value of the best possible defense.

What was the Caroline standard and does it still apply?

The Caroline standard comes from an 1837 incident on the Niagara River, when the American secretary of state Daniel Webster wrote that lawful self-defense required a necessity that was instant and overwhelming, leaving no choice of means and no moment for deliberation. It defined imminence as a threat measured in hours. The standard still appears in legal argument, but in the context of counter-terror operations it has been substantially rebuilt. Modern doctrine treats imminence as a multi-factor judgment that can be satisfied by a continuing pattern of threat or by the last feasible opportunity to act, rather than by a literal clock. Whether that rebuilding is a sensible adaptation or a dissolution of the standard is one of the debate’s sharpest disputes.

What did the Israeli Supreme Court rule in 2006?

In a 2006 judgment, the Israeli Supreme Court sitting as the High Court of Justice declined to declare targeted killing categorically lawful or categorically unlawful. It held that members of terror organizations were civilians who lost protection from attack only insofar as, and only for such time as, they were taking a direct part in hostilities, that each strike had to satisfy proportionality, and that the state had to investigate afterward whether the standard was met. The ruling is the most developed public judicial treatment of the subject by any state. It is also a candid admission that legality can be assessed only case by case, after the fact, which means the standard provides little guidance before a strike is carried out.

What is direct participation in hostilities?

It is the legal concept that determines when a civilian becomes targetable. Civilians are protected from attack, but a civilian who takes a direct part in hostilities forfeits that protection for the duration of the participation. The International Committee of the Red Cross published interpretive guidance in 2009, with the legal scholar Nils Melzer as principal author, proposing that an act counts as direct participation when it is likely to cause harm, has a direct causal link to that harm, and is designed to support one party against another. The guidance also suggested that a person with a continuous combat function in an armed group could be targeted on an ongoing basis. The concept is essential and unstable, because reading it narrowly protects most network leaders and reading it broadly swells the category of the targetable.

Why is imminence so contested in counter-terror operations?

Because the original meaning of imminence, a threat about to materialize, cannot accommodate the way networks plan. A militant commander rarely poses a threat that will materialize within hours. He represents a continuing capacity, with operations planned on timelines of years. Strict imminence would therefore forbid nearly all targeted killing. Defenders of the practice rebuilt the concept around the continuing threat and the last window of opportunity, so that a target who can be located today but not next week is treated as imminent. Critics respond that this rebuilding erases imminence as a discipline, since every senior network figure becomes permanently imminent. Both descriptions are accurate, which is why the dispute does not resolve.

Is killing a militant leader different from killing a uniformed soldier?

Legally, yes, and the difference is the heart of the combatant status problem. A uniformed soldier is a combatant and may be attacked at any time simply by virtue of that status. A militant leader is not a combatant in the legal sense, he wears no uniform and belongs to no state’s armed forces, so by the letter of the categories he is a civilian. He becomes targetable only through the concept of direct participation in hostilities or continuous combat function. This means the lawfulness of killing him depends on a contested characterization of his conduct and role, whereas the lawfulness of killing a soldier depends only on his status. The same man can be a lawful target or a protected civilian depending on which reading a lawyer adopts.

How does the capture versus kill question affect legality?

The necessity principle, sometimes called last resort, holds that lethal force may be used only when no less harmful option is available, which would seem to require capture rather than killing whenever capture is feasible. The entire question turns on the word feasible. Read narrowly, feasible means physically possible, and on that reading killing is rarely necessary. Read broadly, feasible means possible without serious risk to the capturing force and without diplomatic rupture, and on that reading killing is almost always permitted. A state that prefers killing does not need to lie, it only needs to choose the broad reading, and because no one can prove that an unattempted capture would have succeeded, the necessity standard struggles to constrain the choice.

What is the difference between an assassination and a lawful targeted killing?

There is no agreed bright line, and that is precisely the problem the debate keeps circling. The word assassination implies an unlawful, treacherous killing, often of a political figure, and several states maintain domestic prohibitions phrased in those terms. A lawful targeted killing, by contrast, is presented as an act of self-defense or an act of war against a legitimate military objective. Whether a given killing falls on one side or the other depends on the same contested standards examined throughout this analysis. The vocabulary chosen, assassination, elimination, counter-terror operation, neutralization, usually reveals the speaker’s prior conclusion rather than supplying a neutral description.

Does the unwilling or unable doctrine have any limits?

In its current form it has very few, and that is one of the strongest objections to it. Once a state may strike a non-state group on the territory of any host it judges unwilling or unable, the geographic reach of force extends to wherever the group has a presence and the temporal reach extends to however long the group exists. There is no front line and no defined end. The doctrine also lacks a neutral arbiter, the acting state makes the unwilling or unable judgment about another state, by itself, with no agreed threshold. Defenders argue the doctrine is still preferable to a strict reading that would leave victim states without remedy. Critics argue a doctrine without limits cannot meaningfully constrain anything.

Why do critics say targeted killing erodes international law?

The concern is cumulative rather than about any single strike. Each elastic interpretation, of imminence, of feasibility, of direct participation, of the unwilling or unable doctrine, becomes a precedent that the next state can invoke. Critics argue that this normalizes a practice the law was once understood to limit, that it gradually expands the circumstances in which lethal force across borders looks acceptable, and that because powerful states set the precedents, the eroded rules end up serving the strong against the weak. The deeper worry is that a body of law which was built around the boundaries of armed conflict, where and when and against whom force may be used, loses its capacity to constrain once those boundaries are read away.

Does denial of a program count as a legal position?

In a practical sense it does, and India’s case shows how. A state that publishes a legal theory can be measured against it and criticized when an operation fails to fit the stated standard. A state that denies a program entirely offers nothing for the law to grip. There is no admitted policy to scrutinize, no stated test to fall short of, no official act for a court or an inquiry to examine. Denial therefore functions as a legal strategy as much as a diplomatic one, it keeps the operations in a zone where they can be neither defended nor condemned. The structure of the indeterminate law rewards this silence, imposing its costs on states honest enough to acknowledge what they do and sparing the states that refuse.