Daniel's been reading up on targeted assassination programs — the history of how states have hunted down and killed specific individuals they consider threats, outside a battlefield and without a trial. He wants us to trace this through Israel's program, the American one, and the broader historical record going back centuries. The tension he's poking at is this: these are countries with functioning legal systems, not failed states. So how do they square the circle?
The short answer is they don't — they build a parallel legal architecture that runs alongside the courts and operates under different rules. And the thing most people miss is that this isn't some modern innovation. States have been doing versions of this since there were states.
Give me the earliest example you'd actually call state-sponsored assassination, not just battlefield revenge.
The Hashashin — the original Assassins, capital A — in the eleventh and twelfth centuries. They were a Nizari Ismaili Shia sect operating out of mountain fortresses in Persia and Syria, and they developed targeted killing of political and religious leaders into something like a strategic doctrine. They'd infiltrate a target's household over months, sometimes years, then strike in public — often in a mosque on a Friday — to maximize the psychological impact. The Crusaders brought the stories back to Europe, and the word "assassin" entered the Western vocabulary through them.
And the state part of it — they weren't just a cult with knives. They held territory.
Right. They controlled a network of fortified castles, they had a hierarchical command structure, they collected taxes. The Old Man of the Mountain — that's how the Crusaders referred to their leader — was running what amounted to a small theocratic state whose primary instrument of foreign policy was the dagger. Marco Polo wrote about them, though his account is... embellished. He claimed the Old Man drugged young recruits with hashish and took them to a pleasure garden to convince them they'd seen paradise. Hence "hashashin" — hashish-eater. Most historians think that etymology is dubious, but the name stuck.
The drug-addled assassin story is better marketing than "they were very patient and methodical." So that's the medieval template. When does the modern state version start looking recognizable?
The hinge moment is the nineteenth century, and it's not a state doing it — it's anarchists and nationalists using assassination as a political tool, and states reacting. You get the assassinations of Tsar Alexander the Second in 1881, President McKinley in 1901, Archduke Franz Ferdinand in 1914. Those weren't state operations, but they taught states something: a single killing can reshape the political landscape. By the Second World War, you've got the British Special Operations Executive training agents to kill Nazi officers, and the American OSS — precursor to the CIA — running its own operations.
The OSS had a hit list?
They had what they called "the elimination of enemy leaders," though how many they actually got is debated. The more consequential shift comes after the war, when the CIA picks up the baton and starts running assassination as a tool of Cold War statecraft. The Church Committee in 1975 — that's the Senate select committee chaired by Frank Church — documented eight separate CIA assassination plots against foreign leaders between 1960 and 1965. Fidel Castro alone was the target of at least eight attempts. Poisoned cigars, exploding seashells, a contaminated diving suit — the plots got so baroque that they read like farce, but the intent was deadly serious.
The exploding seashell is the one that gets me. You place it where he dives, it detonates, and somehow this is deniable.
The deniability is the whole architecture. The Church Committee found that the CIA had developed a formal protocol for what they called "executive action" — that's the euphemism — with cutouts, intermediaries, and plausible deniability built into every layer. The president would issue a vague authorization, the DCI would pass it down through channels where nobody quite said the word "kill," and by the time it reached the field it was a nod and a wink and a bag of money handed to a contact who knew a guy.
And the legal justification, such as it was?
There wasn't one, really. The Church Committee concluded that the plots violated both international law and domestic law, and in response, President Ford signed Executive Order 11905 in 1976, which explicitly banned "political assassination" by US government employees. Every president since has reaffirmed some version of that ban. Reagan signed Executive Order 12333, which is still on the books, and it states that "no person employed by or acting on behalf of the United States Government shall engage in, or conspire to engage in, assassination."
And yet.
And yet. The ban on assassination has a very specific meaning in American legal doctrine, and it's narrower than most people assume. The executive order prohibits killing for political purposes, but the US government has consistently interpreted "assassination" as excluding killings carried out in self-defense, in armed conflict, or against individuals who pose an imminent threat. So when the US kills someone with a drone strike, the legal framework it invokes isn't assassination — it's the law of armed conflict. The target is classified as a combatant in an ongoing armed conflict, and the strike is characterized as a battlefield action, even if the battlefield is a car on a road in Yemen.
Which brings us to the post-nine-eleven era and the drone program.
The drone program is the most extensive targeted killing campaign any state has ever run. The numbers are staggering. The Bureau of Investigative Journalism has tracked US drone strikes in Pakistan, Yemen, and Somalia since 2002, and their data shows somewhere north of fourteen thousand total strikes, with estimated deaths between nine thousand and sixteen thousand people, of whom between nine hundred and two thousand two hundred were civilians. Those are wide ranges because the US doesn't release comprehensive data, and independent verification on the ground is extremely difficult.
The civilian casualty range being that wide is itself a kind of information. You can't even count the dead precisely.
And the legal architecture that supports it is fascinating in a grim way. The Obama administration developed a two-part framework. First, they asserted that the US was in an armed conflict with Al Qaeda and associated forces, and that this conflict had no geographic boundaries — it followed the target. Second, they developed what became known as the "disposition matrix" — a kill list maintained across multiple agencies, with names added and removed through a process that involved lawyers, intelligence analysts, and military commanders. The president personally signed off on high-value targets.
The "Terror Tuesday" meetings.
That's what they were called internally — weekly meetings in the Situation Room where senior officials reviewed proposed strikes and the president made the final call on who lived and who died. Anwar al-Awlaki, the American citizen killed in Yemen in 2011, was approved through that process. That case is particularly significant because it tested the constitutional limits: can the executive branch kill a US citizen without trial, outside a declared war zone, based on intelligence that is never tested in court?
The Justice Department said yes.
They produced a legal memo — it wasn't released publicly until years later — arguing that due process doesn't necessarily require judicial process when the target is an operational leader of an enemy force who poses an imminent threat and capture isn't feasible. "Imminent" was stretched to mean something closer to "continuing" — the threat doesn't have to be about to happen tomorrow, it's enough that the person is continuously planning attacks. That's a very elastic definition.
That's not a definition, it's a permission slip. What did the memo say about feasibility of capture?
That the executive branch gets to determine it, and the determination isn't subject to independent review. So you've got a closed loop: the same people who decide someone is a threat also decide they can't be captured, and there's no outside check on either judgment. That's the core tension Daniel's asking about — how a rule-of-law state maintains this parallel system without the logic of the parallel system bleeding into the ordinary one.
Let's pivot to Israel. The program there is older than the American drone campaign, and in some ways more institutionalized.
Israel's targeted killing program goes back to before the state itself. In the nineteen forties, the pre-state Jewish underground — the Haganah, the Irgun, and Lehi, which the British called the Stern Gang — all conducted assassinations of British officials and Arab leaders. Lehi in particular was essentially an assassination squad with a political wing. They killed Lord Moyne, the British Minister Resident in the Middle East, in Cairo in 1944. They killed Count Folke Bernadotte, the UN mediator, in Jerusalem in 1948.
Bernadotte is the one that still gets mentioned in diplomatic circles. Killing a UN mediator is a statement.
It is, and the provisional government of Israel arrested hundreds of Lehi members afterward and temporarily suppressed the organization, but most of them were later amnestied. One of Lehi's leaders, Yitzhak Shamir, went on to become prime minister. So from the very beginning, the line between terrorist and statesman was... porous.
That's a generous word for it.
The modern Israeli program really crystallized after the Munich Olympics massacre in 1972. The Mossad launched Operation Wrath of God — that's the campaign to track down and kill everyone involved in planning and carrying out the Munich attack. It ran for years, across multiple countries, and it became the template. The operational details that have come out are remarkable. The Mossad built a dedicated unit called Caesarea, and they used everything from letter bombs to motorcycle-mounted hit teams. They killed Ahmed Bouchikhi in Lillehammer, Norway, in 1973 — except he was the wrong man. He was a Moroccan waiter who looked like the target, Ali Hassan Salameh.
The Lillehammer mistake changed how they operated.
It did. Six Mossad agents were arrested, five were convicted, and the operation was temporarily compromised. After that, Israel tightened its target identification procedures significantly. But they didn't stop. They got Salameh in Beirut in 1979 with a car bomb. They got Khalil al-Wazir — Abu Jihad, Arafat's deputy — in Tunis in 1988. That was a combined naval commando and air force operation, with the team landing on a beach, infiltrating his compound, and killing him in his home.
The Israeli program has a feature the American one doesn't, or at least not in the same way: the legal framework is partly public. They've had Supreme Court rulings on this.
The key case is the 2006 ruling by the Israeli Supreme Court sitting as the High Court of Justice — that's HCJ 769/02, the Public Committee Against Torture in Israel versus the Government of Israel. The court held that targeted killing is not categorically illegal under international law, but it's also not categorically legal. It depends on the circumstances. The court laid out a framework: civilians who take a direct part in hostilities lose their immunity from attack for the duration of that participation, but the determination must be made on an individual basis, with a high standard of proof, and after-the-fact judicial review must be available.
And the "duration of participation" part — that's the same elasticity problem the Americans have. When does a terrorist stop being a terrorist?
The court wrestled with exactly that question and didn't fully resolve it. They said a person who takes a direct part in hostilities on a recurring basis — what they called a "revolving door" pattern — might be targetable continuously, but they didn't draw a bright line. The ruling also requires that if a less harmful means is available — capture rather than killing — it must be used. But like the American framework, the feasibility of capture is assessed by the operational commanders, not by the court.
And the court reviews it afterward, not before.
Correct. The review is ex post, and it's limited. The court said it would examine whether the decision fell within a "zone of reasonableness" — that's a very deferential standard. In practice, no targeted killing has ever been ruled illegal by the Israeli Supreme Court. The framework exists, it looks like law, but the teeth are... well, they're not really teeth.
The numbers on the Israeli program — what do we know?
B'Tselem, the Israeli human rights organization, has documented that between the start of the Second Intifada in September 2000 and the end of 2023, Israel carried out targeted killings that resulted in the deaths of at least four hundred sixty-two targeted individuals and at least three hundred twelve bystanders. Those are confirmed cases. The actual numbers are probably higher, especially during the Gaza operations where documentation is harder. During Operation Protective Edge in 2014, B'Tselem documented strikes on thirty-one homes of suspected militants that killed one hundred eight people who were not the targets, including forty-one children.
The house-strike tactic is a whole separate category of escalation.
It is, and it's where the Israeli program diverges most sharply from the American one. The US drone program, for all its flaws, generally uses relatively small munitions and the strikes are aimed at individuals in vehicles or on foot. Israel has repeatedly used one-ton bombs on residential buildings to kill a single target, knowing that everyone else in the building will die. The operational logic is that a building can be reconstructed, but the intelligence opportunity to get the target may not come again. The legal logic is that the civilian casualties are proportionate to the military advantage of eliminating the target.
It always does. Proportionality in the law of armed conflict doesn't mean the number of civilians killed must be less than the number of combatants. It means the expected civilian harm must not be excessive in relation to the concrete and direct military advantage anticipated. That's a balancing test that is inherently subjective, and the people doing the balancing are the ones pulling the trigger.
Let's go broader — you mentioned the historical arc earlier. What does the pre-modern record look like beyond the Hashashin?
Assassination as a tool of statecraft is as old as states. The Roman Republic had a formal institution called proscription — Sulla posted lists of enemies of the state in the Forum, and anyone on the list could be killed with impunity, with the killer receiving a share of the victim's property. Cicero was killed under the proscriptions of the Second Triumvirate in 43 BCE. His head and hands were displayed on the Rostra in the Forum.
That's not even assassination in the shadows. That's assassination as public spectacle with a property incentive.
The Byzantines refined it into something closer to a bureaucratic function. They had a formal office of political murder — the agents were called the "silentiarii" in some accounts, though the terminology varies. They used poison, strangulation, and blinding as standard tools of succession management. Blinding was particularly Byzantine — it rendered a rival politically dead without the religious complications of actual murder, since blinding didn't technically kill the person.
A theological loophole for mutilation. That's a level of legalism I almost admire.
The Venetians ran what was probably the most institutionalized assassination program in European history. The Council of Ten — that's the secretive ruling body of the Venetian Republic — maintained a network of assassins and poisoners, and they kept formal records of their decisions. If the Council voted to eliminate someone, the state poisoner — the official title was something like "master of poisons" — would be dispatched. The records show they debated these decisions in writing, with votes recorded, as though it were a trade policy.
The bureaucracy of murder. There's something almost more chilling about the written vote than the killing itself.
The Ottomans had their own version. The practice of fratricide upon succession — a new sultan killing all his brothers to prevent civil war — was actually codified in law by Mehmed the Conqueror in the fifteenth century. He wrote: "To whichever of my sons the sultanate shall be granted, it is appropriate that he kill his brothers for the good order of the world." That's not assassination of foreign enemies — it's institutionalized murder within the ruling family, justified as state necessity.
And the Mughals, the Chinese empires — everyone had some version of this.
The Chinese imperial system had a particularly refined version: the "grant of death" — ci si, death by imperial decree, often carried out with a silk cord or a cup of poisoned wine delivered to the disgraced official's home. The official was expected to accept it, thank the emperor for the honor, and carry out the sentence himself. It preserved the fiction that the emperor wasn't killing anyone — he was merely indicating his preference, and the subject was choosing to comply.
That's the same deniability impulse in a different cultural idiom. The emperor doesn't kill, he merely... suggests.
And you can draw a straight line from the silk cord to the drone strike authorization memo. The technology changes, the legal language gets more sophisticated, but the fundamental move is the same: the state wants someone dead, and it constructs a procedural framework that lets it happen without anyone having to say "I ordered this man killed."
What about the argument that targeted killing is actually more restrained than the alternative? The people who defend these programs say: we could bomb a training camp and kill fifty people to get one target, or we can put a missile through his car window and kill three. Isn't the targeted approach the more ethical one?
That argument has real weight, and it's not just a talking point. When the US was fighting ISIS in Mosul, the alternative to targeted strikes was urban artillery bombardment, and that killed civilians on a vastly larger scale. The precision strike is less destructive than the conventional alternative in many scenarios. The problem is that the availability of the precision option also lowers the threshold for using lethal force. When the only way to kill a target was to send in a commando team, you did it rarely because the risk to your own forces was high. When you can do it with a drone operated from Nevada, the political cost drops to near zero.
So the tool makes the decision easier, and you get more decisions.
And more decisions means more mistakes, and more mistakes means more people who were in the wrong place at the wrong time, and those people have families who remember. The strategic question that never gets answered satisfactorily is whether the long-term recruiting benefit to the enemy — from every wedding party that gets hit, every wrong house — outweighs the tactical gain of removing one operator.
The tactical gain is measurable. You can count the attacks the dead man didn't plan. The strategic cost is diffuse and unfolds over years.
And that asymmetry makes the cost-benefit analysis inherently biased toward action. The intelligence agency can point to a specific threat eliminated. Nobody can point to the specific recruit who joined because his cousin was killed in a strike and say "that recruit wouldn't exist if we hadn't pulled the trigger." It's the classic problem of measuring a counterfactual.
Israel has a term for this — "mowing the grass." You're not solving the problem, you're managing it at an acceptable level of violence.
That phrase originated in Israeli strategic thinking about Gaza specifically. The idea is that you can't eliminate the threat permanently — the underlying conditions that produce militants will persist — but you can periodically degrade their capabilities through targeted strikes, keeping the threat below a threshold where it threatens daily life. It's a deeply pessimistic doctrine. It assumes the conflict is permanent and the goal is management, not resolution.
And the targeted killing program is the precision instrument of that management strategy.
Right. Which is why the legal frameworks in both the US and Israel are designed for permanence, not for exceptional circumstances. They're not emergency measures that expire when the crisis passes. They're institutional architectures built to operate indefinitely.
The thing I keep coming back to is the asymmetry of information. The state knows why it's killing someone. The family of the dead man doesn't get to see the intelligence. The public doesn't get to see it. The courts either don't see it or see only what the state chooses to show them. The entire system runs on trust that the intelligence is good, and there's no mechanism for falsifying that trust.
And the track record on intelligence quality is... mixed. The Lillehammer mistake is the famous one, but there are plenty of others. In 2002, Israel dropped a one-ton bomb on a building in Gaza City to kill Salah Shehadeh, a Hamas military commander. They got him, but they also killed fourteen other people, including nine children. The operational judgment was that the intelligence on his location was solid and the window was closing. The intelligence was right about the location. The question is whether the proportionality assessment — fourteen lives for one — was right, and there's no way to litigate that after the fact.
The Shehadeh strike is the one that led to the Supreme Court case you mentioned earlier.
It was one of the catalysts, yes. The petitioners argued that the strike was a war crime. The court didn't rule on that specific strike — it used the case to establish the general framework I described. And the framework effectively said: these decisions are justiciable, meaning courts can review them, but the standard of review is so deferential that the operational judgment almost always stands.
There's a phrase for that in legal scholarship — "the appearance of legality without the substance of constraint."
That's exactly the critique. And it applies to the American system too. The Obama administration was very careful to build a legal architecture that looked robust — the memos, the interagency review, the presidential sign-off. But the architecture was designed to authorize, not to constrain. The number of proposed strikes that were rejected at the presidential level was vanishingly small.
Do we know the number?
Not precisely. Some former officials have said that Obama rejected more strikes than he approved, but they've never provided numbers, and the process is entirely classified. We have their word for it, and their word is the only evidence.
Let me ask you the normative question Daniel is really getting at. Is there a version of targeted killing that is compatible with the rule of law? Or is the whole enterprise inherently lawless, and the legal frameworks are just window dressing?
I think there are versions that are more and less compatible. The Israeli Supreme Court framework, for all its flaws, at least establishes that these decisions are subject to legal standards and judicial review. That's not nothing. In a world where states are going to conduct targeted killings regardless — and they are — having a court articulate limiting principles is better than having no principles at all.
But the limiting principles don't limit very much.
They limit at the margins. They probably prevent the most egregious cases — the purely personal vendetta, the killing of someone who's clearly not a legitimate target. But they don't prevent the hard cases where the intelligence is ambiguous and the civilian harm is high. Those are the cases that matter, and those are the cases where the legal framework effectively defers to the operational judgment.
The American program has an additional problem: it's entirely executive. Congress has never authorized it. The 2001 Authorization for Use of Military Force is the statutory hook, and it's been stretched to cover groups that didn't exist in 2001, in countries that weren't mentioned, against individuals who weren't born when it was passed.
The AUMF is now twenty-five years old and has been used to justify military action in at least nineteen countries. That's not a legal framework — that's a blank check with a date stamp. And Congress has repeatedly declined to pass a new authorization, which means the executive branch operates in a zone of constructive legislative acquiescence. Congress doesn't authorize, but it also doesn't prohibit, and it funds the operations. That's a political choice, not a legal vacuum.
The political choice is the key. These programs persist because they're politically popular, or at least not politically costly. The American public doesn't feel the drone war. The Israeli public feels the threat of terrorism directly and broadly supports the targeted killing program. The legal debates happen in courtrooms and law reviews, but the political support is what keeps the drones flying.
And the political support is sustained by the fact that these programs are, in a narrow tactical sense, effective. They do kill the people they're aimed at. The question is whether tactical effectiveness translates into strategic success, and on that the evidence is much murkier. Did killing Anwar al-Awlaki degrade Al Qaeda in the Arabian Peninsula? Temporarily, probably. Did it eliminate the threat? AQAP is still operating. Did the civilian casualties from the broader drone campaign in Yemen create more militants than they killed? We don't know, and the people running the program have no incentive to find out.
There's a term from medicine that applies here — iatrogenic harm. Harm caused by the treatment itself.
And in medicine, we have randomized controlled trials to measure iatrogenic harm. In counterterrorism, we have classified assessments produced by the same agencies conducting the operations. The epistemic problem is fundamental.
So where does that leave us? We've got a practice that's ancient, that every major power has engaged in, that's legally ambiguous at best, that's tactically effective and strategically uncertain, and that shows no sign of stopping.
I think it leaves us with the recognition that the rule of law, in this domain, is aspirational rather than operational. The legal frameworks are real — the Supreme Court ruling, the executive orders, the memos — but they function more as boundary markers than as genuine constraints. They tell the state: you can do this, but you must do it in this way, with this process, and you must be able to explain afterward why it was necessary. That's not nothing. But it's also not the same thing as being bound by law in the way an ordinary criminal defendant is bound.
The process is the punishment, except in reverse. The process is the permission.
Hilbert: Nineteen ninety-one. I was driving a delivery van for a printing company in Newark. One of the press operators — man named Kovacs, Hungarian, had been in the country about six years — came in one morning and said his brother-in-law had been killed in Bucharest during the revolution. Not in the fighting. Afterward. Someone put his name on a list, and two days later he was pulled out of his apartment and shot in a stairwell. Kovacs said the list wasn't official. It was a committee of former dissidents who'd taken over the security ministry and were settling scores. He said "they killed him because his neighbor didn't like him and told the committee he was Securitate."
A personal grudge routed through revolutionary justice.
Hilbert: That's what Kovacs said. The neighbor wanted his apartment. The committee needed names to show they were purging the old regime. Everyone got what they wanted except the dead man. I thought about that when you were talking about the disposition matrix. The American version has more lawyers. But the dynamic where someone who isn't accountable puts a name on a list and that name becomes a death sentence — the shape of it is the same. The lawyers make it feel different, but at the receiving end, a missile doesn't care whether a judge reviewed the file.
The Kovacs story is the endpoint of the asymmetry I was describing. The target doesn't know he's on a list. He doesn't know what evidence was presented. He doesn't get to argue back. He just dies, and afterward someone says the process was rigorous.
Hilbert: The other thing Kovacs said — he was angry for years, but what really ate at him wasn't that his brother-in-law was killed. It was that there was no record. Nobody would say who put the name on the list. Nobody would say what the accusation was. The committee dissolved, the files disappeared, and the neighbor still lives in the apartment. He said "I don't even know what I'm supposed to forgive."
That's the closure problem. One of the arguments for judicial process — real judicial process, not ex post review — is that it produces a record. The state has to say, publicly, "we did this, for these reasons, based on this evidence." The targeted killing frameworks in both the US and Israel are designed to avoid producing that record. The intelligence stays classified, the deliberations stay secret, and the families get no accounting.
Hilbert: The van I was driving was a Ford Econoline. Three hundred twelve thousand miles on it when the transmission went. I ever tell you about the transmission?
You haven't.
Hilbert: Rebuilt unit from a junkyard in Elizabeth. Cost me four hundred dollars. The mechanic said it would last six months. It lasted four years. I sold the van to a guy who was going to drive it to Florida. He called me from South Carolina to say the transmission was still fine. Sometimes the thing that's supposed to fail doesn't.
I'm trying to decide if that's a metaphor for the targeted killing programs or the opposite of one.
Hilbert: It's just a transmission. But I think about it when people say these programs are unsustainable. Things that shouldn't work sometimes work for a long time. The question isn't whether they'll fail. It's who's in the stairwell when they do.
If you take one thing from this, it's that the legal frameworks around targeted killing are real but they're designed to authorize, not to constrain. The distance between "reviewed by lawyers" and "constrained by law" is the whole ballgame.
And the history tells us this isn't a bug — it's the feature. From the silk cord to the drone strike, the goal has always been to kill the target while maintaining the appearance of lawful process. The innovation of the modern era is that the appearance has gotten much more sophisticated. Whether the substance has changed is a harder question.
We didn't even get to the Russian program, the British program in Northern Ireland, the French in Algeria. There's a lot more history here.
The French program during the Battle of Algiers is its own episode. The use of torture to locate targets, the death squads, the way it poisoned French politics for a generation. But that's a different conversation.
We'll put it on the list. Thanks to our producer Hilbert Flumingtop for keeping us honest.
This has been My Weird Prompts. If you want to send us your own weird prompt, you can reach us at show at my weird prompts dot com, or find everything at my weird prompts dot com.
We'll be back soon.
