The Law and Tech of Two Targeted Killings

The New York Times recently published two fascinating pieces on two separate instances of targeted killings. The first is on the tragic denouement of the 20 years of US presence in Afghanistan – a drone strike conducted on 29 August by the US military in Kabul, purportedly against terrorists planning a second deadly attack against the international airport there. Instead of terrorists, however, the drone strike killed 10 civilians, including seven children; Zemari Ahmadi, the driver of the car struck by the drone was in fact an aid worker for a US NGO, who had hoped to seek asylum in the United States. The US military admitted its error after an investigation by the Times and journalists on the ground.

The second piece is a detailed exposé of the 27 November 2020 assassination by Israeli agents of Mohsen Fakhrizadeh, the scientist leading the Iranian nuclear programme. This was the culmination of a series of such targeted killings of scientists involved in the Iranian nuclear programme, and its mode of execution – the use of an AI-assisted and remotely controlled machine gun in a roadside ambush while Fakhrizadeh was driving his car to his country house near Tehran – is straight from the pages of a sci-fi novel. Indeed, when Iranian authorities first came out with this account of Fakhrizadeh’s killing the story was regarded as so fanciful that it was widely disbelieved, both in Iran and abroad. But it turns out that a killer robot – albeit not a fully autonomous one – was in fact used to assassinate him; the Times story is very much worth the read.

There are several interesting parallels to be drawn between these two cases. One is the use of technology in implementing the assassinations, and in particular in ensuring – or not – that the target is the correct, intended one. For example, US drone operators were able to follow Mr Ahmadi’s activities for some eight hours before launching the missile strike. Yet despite this capability every new item of information suffered from biased processing on the basis of initial faulty intelligence, leading to one mistaken factual judgment after another – and so an aid worker loading water canisters into his car was perceived as a terrorist loading explosives intended for the Kabul airport. On the other hand, Israeli agents were more successful. Not only were they able to follow Mr Fakhrizadeh’s movements with incredible precision so as to set up their ambush, but they deployed a machine gun, controlled via satellite by a remote human operator outside Iran, assisted by AI which enabled the weapon to take into account the time-lag between the operator’s input and the gun firing, to properly adjust for the gun’s recoil, and to use facial recognition to avoid hitting Mr Fakhrizadeh’s wife, sitting next to him in the front passenger seat of the car, or other bystanders. This was an incredibly sophisticated machine, smuggled in parts into Iran and then carefully assembled and calibrated by Israel’s human assets in the field, who however did not need to be nearby when the assassination itself took place. The only technology failure was that a bomb that was meant to destroy the gun after the assassination did not do so fully.

Then there is the use of technology by external actors trying to establish the truth of what had actually happened, something we are becoming increasingly used to in the age of open-sourced investigations. And, finally, there is the law – both cases suffer from substantial uncertainty as to what sub-regimes of international law even apply to the given use of lethal force, and then onwards from uncertainty as to how they apply. The rest of the post will focus on these legal questions of the jus ad bellum, the jus in bello, and human rights law. I will not be dealing with any issues of US and Israeli law.

The law on the use of force/jus ad bellum

The first issue that arises in both cases is whether the targeted killings by US and Israel constituted uses of force against Afghanistan and Iran respectively, which would be prima facie prohibited under Article 2(4) of the Charter. The principal problem here is whether relatively low-level use of lethal force by state agents – the killing of a single individual, if Fakhrizadeh’s death was looked at in isolation, or of several individuals over a more extended period of time (if we take into account other assassination operations by Israel), or of 10 individuals in a single moment in time (which is the case with the botched US drone strike) – count as ‘force’ for the purpose of Article 2(4). And if either killing was a use of force in that sense, a further question would be whether they qualify as ‘armed attacks’ under Article 51 of the Charter, requiring an additional gravity element and triggering the right to self-defence. These questions were explored, for example, in the very insightful posts by Tom Ruys on Just Security and by Dapo here on EJIL: Talk! in the context of the Salisbury novichok attacks, so I will not belabour them here. Suffice it to say that the killing of a single individual by one state on the territory of another can quite reasonably be said to constitute a use of force against the latter, and that this assessment only become easier when the wider context is taken into account.

The Israeli and American operations can thus both be regarded as uses of force that would be contrary to the UN Charter in the absence of a lawful ad bellum justification. The US drone strike can potentially be justified on two different bases – consent and self-defence. As for the former, the US could argue that its military presence in Afghanistan was consented to by the government of Afghanistan, and that this consent encompassed operations such as this one. The problem here is that the government that provided this consent had fallen by the time the drone strike took place. It is unclear whether at the relevant time the state of Afghanistan had no government to speak of, or whether the Taliban should already be regarded as such a government (if they so qualify even today). Put differently, did the consent provided to the US lapse with the demise of the Ghani government? Even if it did lapse, is there some reasonable period of time during which US forces would have to withdraw but during which their continued presence would remain lawful? Even so, could, perhaps, the Taliban’s cooperation with the US regarding the protection of the Kabul airport be regarded as some form of implied consent?

The uncertainty surrounding any argument about consent would likely mean that the US would prefer to rely on a self-defence justification for its drone strike – viz., that striking ISIS terrorists was necessary to stop them from attacking US and allied forces at the Kabul airport. But that argument is also hardly straightforward. First, it assumes that the US presence at the airport was lawful, which leads us back to the consent issue examined above – a state cannot invoke self-defence to protect its forces deployed in another state if that deployment is itself unlawful.

Second, and more importantly, the people actually struck by the US were innocents, not ISIS terrorists intending to attack the airport. In other words, the US strike was tainted by a mistake of fact, and the question of principle here would be whether a subjectively honest and objectively reasonable such mistake would suffice to excuse the US of responsibility. I will refer readers to my series of posts on mistakes of fact in the use of lethal force for a detailed discussion (here, here and here). Very briefly, a subjectively honest mistake – or, as US General McKenzie put it, that the drone operators were acting “in the profound belief” that they were targeting terrorists – would clearly not suffice to excuse the US from liability. But an objectively reasonable mistake might.

While the US may have some plausible argument to justify its erroneous strike under the jus ad bellum, Israel has no such argument for killing Mr Fakhrizadeh and other Iranian nuclear scientists. It could only make a theoretical claim to self-defence – that killing F was necessary to stop Iran from developing a nuclear weapon that would be used against Israel – but that claim is unsustainable even under expansionist contemporary approaches to self-defence, such as those of the US and the UK.  One common line of argument here is that self-defence is permissible against an anticipated, future armed attack only if that attack is imminent. I have discussed this notion of imminence previously on the blog, e.g. in the context of the Soleimani strike (see here and here).

If ‘imminence’ was construed in temporal terms, there clearly is no such imminent nuclear attack by Iran against Israel that killing F could disrupt. But even if ‘imminence’ was construed, as some expansionist accounts do, in purely causal terms so that an imminent armed attack was one where the victim of that future attack had to act now in order to stop it, no such attack is evident here. Even if we were to accept that Iranian leadership was irrevocably committed to developing a nuclear weapon (which they may or may not be), there is zero evidence that they are irrevocably committed to striking Israel with that weapon. Indeed such a course of action would be suicidal for the Iranian regime, if for no other reason then because Israel already has its own nuclear arsenal. It therefore cannot be necessary for Israel to act now to prevent that attack from happening. Thus, even if the notion of imminence was taken at its broadest there is no plausible way of arguing that Iran will commit an imminent nuclear strike against Israel, and that killing F was necessary to stop this.

Obviously, Israel perceives a nuclear Iran as an existential threat, and is willing to act now to reduce that threat. But this mode of thinking is purely pre-emptive – it simply cannot be described as self-defence, any more than some of the expansive doctrines of the first George W. Bush administration could be so described. The potential magnitude of the threat does not alter this position. Israel already has a nuclear capability that can devastate Iran, yet we would not accept that Iran can act today against Israel in some putative form of ‘self-defence.’ The United States and Russia have the capability to destroy each other many times over, yet even during the height of the Cold War we would not have accepted that the mere possession of such a capability and the adversarial posture of the two states sufficed to trigger a claim of self-defence. Accepting that view would mean that any state would be entitled to act forcibly against any other state that it perceived as a threat.

International humanitarian law/jus in bello

Moving now to IHL, both targeted killings raise interesting questions – especially of IHL’s applicability. Turning first to Fakhrizadeh’s assassination, one view would be that he was killed in peacetime, that his killing had no nexus to any armed conflict, and that therefore IHL simply did not apply. An alternative view could be that at the time of the strike there was a pre-existing international armed conflict (IAC) between Israel and Iran, for example due to hostilities between Israel and Iranian armed forces and proxies in Syria. But even if this latter view was taken as factually and legally correct, this would not obviate the need for a jus ad bellum analysis as above – the existence of an IAC would not entitle Israel to initiate an all-out assault against Iran on Iran’s own territory, and vice versa.

Let’s say that there was in fact a pre-existing IAC between Israel and Iran, or that the AI-assisted assassination created such an IAC, which lasted only very briefly. Would F have been a lawful military target in that IAC – would, in other words, he be classified as a combatant?

It is entirely straightforward that scientists, even if engaged in research with a clear military application, do not thereby become combatants. A facility they work in may be a lawful military objective if by its nature, location, purpose or use it makes an effective contribution to military action and if its partial or total destruction, in the circumstances ruling at the time, offers a definite military advantage (Rule 8, ICRC Customary IHL Study). A scientist working in that facility could be lawfully killed incidentally if the principle of proportionality is complied with. But a scientist does not stop being a civilian simply because their work contributes to the military effort, no more than a factory or farm worker whose labour sustains the war effort.

Yet Fakhrizadeh’s personal situation may not be all that simple. While he was undoubtedly a scientist and a university professor, he also held the rank of a brigadier general in the Iranian Revolutionary Guards. The Guards can easily be regarded as part of Iranian armed forces; in particular, although it is a very complex and multifaceted organization, Article 150 of the Iranian Constitution clearly sees the Guards as part of the Iranian armed forces. F could therefore be a combatant as a member of that armed force; only medical and religious personnel of the armed forces are excluded from the definition of a combatant (Rule 3, ICRC Study), and he wasn’t either. In short, his position may be no different from Qasem Soleimani, also a general in the Revolutionary Guard, and if that is right killing him would not be a violation of IHL, to the extent IHL even applied – but again, this would not obviate the need for a jus ad bellum analysis, as I have explained above.

Turning now to the US drone strike in Kabul, we have another novel applicability issue. There is no doubt that prior to the fall of Kabul the US was engaged in a non-international armed conflict (NIAC) against the Taliban, fighting them jointly with the then-Afghan government. Let’s leave aside for the moment the problem of the continued existence of the Afghan government and the impact thereof on the existence and classification of the conflict. The bigger issue is that the drone strike was not even putatively directed against the Taliban – indeed the US was at the time actively cooperating with the Taliban, including through the sharing of intelligence, in order to protect the evacuation effort at the airport. Rather, the strike was directed against suspected terrorists hostile to the Taliban belonging to ISIS-Khorasan Province, even if the people ultimately killed had nothing to do with ISIS-K. I thus fail to see how the drone strike would have a nexus to the US/Taliban NIAC in Afghanistan. The strike would be regulated by IHL only if ‘protracted armed violence’ existed between US forces and ISIS-K, i.e. if there was a discrete NIAC between the US and this armed group, which does not seem to be the case.

If IHL did apply to the US drone strike, the substantive question would again be one of mistake of fact in targeting. As I explained in my previous series of posts, the position under IHL would arguably be that honest and reasonable mistakes of fact as to distinction and proportionality – i.e. those that relate to the identification of the target and to anticipated civilian losses – could excuse the state. But this will only be the case if all reasonable precautions were taken, and in particular if all feasible measures were taken to verify that the target was a military objective (Rule 16, ICRC Study). This is primarily a factual question on which I hold no firm view, but the account given in the New York Times article would not seem to very supportive of the reasonableness of the error made, however honest it may have been. Consider, for example, the supposed reliance on the model of the car driven by Mr Ahmadi as an indicator of his hostile intent, when that model is very commonly used in Kabul.

Human rights law

International human rights law is arguably a much better fit for assessing the legality of the two targeted killings than IHL. And even if the two killings were ad bellum compliant this would not ipso facto make them consistent with the human right to life. If, for example, the US clearly had the consent of the Afghan government to undertake its operation, it would still need to provide a justification for using lethal force against the individuals whom it targeted.

But there is an applicability issue here as well – the problem of extraterritoriality, because both strikes took place outside the territories of the states using lethal force. The United States and Israel have long opposed the extraterritorial applicability of human rights law, and they have done so precisely in order to be able to engage in operations such as these with fewer legal constraints. But that is in the long run a losing proposition; there is simply no satisfactory normative argument that justifies excluding extraterritorial assassinations from the scope of human rights protection.

I discussed this question extensively last week in light of the Carter judgment of the European Court of Human Rights, in which it found that the ECHR did apply to the killing of Alexander Litvinenko by Russian agents in London. Readers will recall that the Court rightly adopted an expansive approach to extraterritorial jurisdiction but that in doing so it also relied on the dubious notion of ‘proximate targeting’, borrowed from its judgment Georgia v. Russia No. 2. Thus, a person ‘proximately targeted’ (whatever that means) would be entitled to the protection of their right to life extraterritorially, whereas a person killed by state agents but not so targeted would not be. The arbitrariness of that position (which the Chamber really used to undermine the restrictive approach in Bankovic and GvR No. 2) is well exposed by the two cases before us. Is a person killed by a drone, whose movements were carefully observed for some 8 hours by the drone operators, someone who was ‘proximately targeted’? Is a person killed by a robotic machine gun, but one operated remotely via satellite from hundreds of miles away and with a smidgeon of AI assistance, someone who was ‘proximately targeted’? Simply put, no sensible system of human rights protection could regard extraterritorial killing by poisoning to be covered by the ECHR or some other human rights treaty, but for the US drone strike or the Israeli super-machine gun assassination not to be so covered.

Which brings us to the merits of the potential right to life violation. The analysis here is not too dissimilar from the jus ad bellum. Again Israel would to me seem to have no colourable justification here. A lethal force against an individual could be justified under IHRL only if they posed an immediate threat to the lives of others, and no other options were available to deal with that threat. But Mr Fakhrizadeh posed no such threat, despite his involvement in the Iranian nuclear programme. Human rights law simply cannot justify the killing of any scientist whose work may, at some point in the distant future, be used to inflict harm on others. Otherwise we could just declare an open hunting season on university professors and government scientists worldwide.

As for the US drone strike, the issue here would again be one of mistake of fact. Under the principles articulated in McCann a mistake that is both honest and reasonable – in the sense that all feasible precautions were taken in the planning of the operation to minimize the loss of life – would excuse the state from a violation of the right to life, even if the perceived threat to lives of others was not really there. Whether this is the case here or not would again depend on a deeper analysis of the facts, including the intelligence information the US operators had available at the time, but as things stand it seems unlikely that the US could successfully discharge this burden.

In short, the erroneous US drone strike in Afghanistan most probably violated international law (depending on the resolution of certain questions of fact), and Israel’s killing of Mr Fakhrizadeh very definitely did so (we know all the facts we need to know). The US will likely make an ex gratia payment to the victims of the strike without acknowledging its legal responsibility to do so, as it sometimes does, and is conducting an investigation into how the error was made; Israel will do no such thing, as the killing was approved at the highest political level. I do hope to have shown, however, that engaging with human rights law is absolutely necessary to properly evaluate extraordinary cases such as these.