The following August, the Russian opposition leader Alexei Navalny became violently ill on a flight. After doctors placed him in an induced coma, it soon became apparent that someone had poisoned him with a deadly nerve agent, Novichok, the very same substance used in earlier attempts to assassinate other Russian dissidents and defectors. Navalny survived, and, in a remarkable twist, later posed as a senior Russian official and telephoned one of the assassins to demand why the poisoning had failed. Through subterfuge of his own, he learnt that they had applied the poison to his underpants in a hotel room, but that it had failed to penetrate his skin as planned.

In November, just three months later, Mohsen Fakhrizadeh was ambushed, shot and killed outside Tehran. This was no ordinary murder. Fakhrizadeh was Iran’s top military nuclear scientist, and, rather mysteriously, the gunmen were nowhere to be seen. His heavily armed bodyguards could only haplessly shoot back into thin air. The murder weapon – a robotic machine gun – was equipped with artificial intelligence and controlled by satellite to target Fakhrizadeh, and only Fakhrizadeh. His wife, sitting centimetres away from him in the car, escaped unharmed. A stray dog wandered into the line of fire; the bullets missed. The attack was remarkably precise, leaving no collateral damage. Iran quickly accused Israel of assassination. These three high-profile attacks – two of which succeeded – spanned the year 2020. Each was conducted by a different state using different methods. Whether they constituted assassinations or targeted killings is debatable, but one thing is clear: state killing of high-profile targets remains a feature of international politics today.

Over the years, the Americans have turned to poisoned toothpaste, poison-lined scuba diving suits and, most famously of all, exploding cigars. Meanwhile, the Israelis have launched over two thousand assassination operations. Methods include snipers, car bombs, parcel bombs, explosives hidden in a phone, and poison.¹ They were not always successful. On one infamous occasion, back in 1997, Mossad operatives travelled to Jordan on a mission to assassinate the leader of Hamas, the Palestinian militant group. The top-secret plan was to insert poison into his ear, but it went spectacularly wrong. The perpetrators were caught and, as the target fell dangerously ill, Israel had to provide the antidote as a diplomatic row broke out.

There may be a thousand ways to die, but there are only really two ways for states to kill: directly or indirectly. States can use their own forces – usually paramilitary, intelligence or special forces – to take out the target. In the examples cited above, the US, Russia and Israel did just that. Soleimani was killed by a US drone; Russian intelligence poisoned Navalny; and Israeli intelligence organized the assassination outside Tehran, smuggling ground-breaking technology into the country. In fact, Russia passed a law in 2006 permitting its military and special services to conduct just such extrajudicial killings abroad, targeting those accused of ‘extremism’ and ‘hooliganism’.²

A direct strike has the benefit of greater control over the operation and a higher chance of success. State assets are well trained, equipped and funded. But there is a catch. Greater control comes at the cost of decreased plausible deniability. Direct involvement, getting hands dirty, will more often than not leave the state’s fingerprints all over the operation, making it much more difficult to deny. Who else other than Israel could have set up a remote-control semi-autonomous gun amid the villas of an upmarket district outside Tehran? It hardly looked like an accident or the work of a random run-of-the-mill terrorist. Given the risks of exposure, states are more likely to kill directly if they deem it legal, if they are not too concerned about secrecy, or, as we shall see, if they think the benefits outweigh the negatives.

Presidents and prime ministers contemplating a lethal strike grapple with this macabre trade-off between directness and deniability.³ Drones and special forces sit at one end. Next, a spymaster might recruit an agent to carry out the assassination, thereby giving some deniability. In 2013, a young Turk living in France shot dead three Kurdish activists in Paris. One was Sakine Cansız, a political refugee and co-founder of the PKK terrorist group, established to fight for Kurdish independence from Turkey. Press reports strongly suggested that Turkish intelligence had recruited the assassin, but it was difficult to prove.⁴

The agents could even be unwitting. In 2017, North Korea assassinated the dictator’s half-brother, Kim Jong-nam, in a bizarre operation. Jong-nam had arrived at Kuala Lumpur airport, travelling under a fake name, on his way back from a picturesque tourist island off Malaysia’s west coast. He was a wanted man, having become a critic of the regime – and a reported CIA source. Indeed, he may have been travelling back from meeting an American intelligence officer. Kim waited in the busy terminal for a flight on to Macau when, all of a sudden, a woman swooped in behind him thrusting a wet cloth into his face. His eyes burned; he felt dizzy. The liquid was a deadly nerve agent and Kim died shortly afterwards.

A twist followed. When arrested, the woman and her accomplice claimed that they had no idea what they were doing, instead believing it was a prank for a hidden-camera television show. They had earlier been seen performing similar pranks in a nearby mall. The supposed TV producers who had recruited them quietly fled the country.

Moving further down the macabre scale of directness and deniability, an intelligence agency sometimes supplies a rebel group with weapons – but issue no orders and ask no questions. The CIA has a track record here. Back in the 1950s, it covertly supplied pistols, ammunition and grenades to dissidents seeking to overthrow Rafael Trujillo in the Dominican Republic. Trujillo was a thug of a dictator, amassing a personal fortune while unleashing terrorism and massacre at home and abroad. Rumour has it he fed the bodies of his dead opponents to sharks. His cult of personality – visible through a superabundance of statues and even the renaming of the highest mountain in the country after himself – made him a perfect target for assassination. If he fell, so too would the entire regime.

Armed with weapons and ammunition from the CIA, but without explicit American instruction, conspirators ambushed Trujillo’s car along a dark highway in May 1961. A vicious gun battle left the dictator sprawled dead on the road. The US did not technically order the assassination, but some in Washington knew it was coming and fully approved.⁵

Twenty years later, with assassination now illegal under US law, the CIA covertly supported rebels in Nicaragua seeking to overthrow the leftist government. As part of the operation, the CIA famously prepared a training manual which highlighted what it obliquely called the benefits of selective violence for propaganda purposes. In other words, assassination. Critics decried it as a murder manual. At around the same time, CIA officers covertly supporting the Afghan mujahideen against the Soviet invasion sent sniper rifles and other lethal weapons to the rebels – but asked no questions about how they would be used.⁶

If special forces and drones lie at one end, then collusion lies at the other end of our directness/deniability scale. This might even be as passive as simply turning a blind eye to terrorist activity or not investigating crimes properly. Collusion is particularly associated with UK activity in Northern Ireland; however, it extends well beyond this arena. Loyalists and collaborators played an important role in the rise and fall of the British Empire, from surrogates in nineteenth-century India killing on behalf of the crown, to collusion with friendly local forces during dirty end-of-empire counter-insurgencies in places like Palestine, Cyprus and Kenya.

As part of the covert operation to overthrow the Iranian prime minister in 1953, MI6 used propaganda to smear Iran’s chief of police and whip opposition forces into a fervour. He was kidnapped (as part of the plan to provide a morale boost for the opposition) but ended up tortured and murdered. Two years later, MI6 tried to start a row within the Hanoi politburo intended to spark a chain of events leading to the assassination of the North Vietnamese president Ho Chi Minh. No order had to be given; the death would not be traced back to Britain. In 1960, Britain ran a covert campaign to destabilize the Congolese leader Patrice Lumumba. It included propaganda, bribery, smears and even use of stink bombs. Alongside the CIA, MI6 conspired to create a situation whereby the death of Lumumba was practically inevitable.⁷

Collusion in Northern Ireland led to the deaths of numerous republicans. Most notoriously, in 1989, loyalist terrorists burst into the home of a thirty-nine-year-old nationalist solicitor, Pat Finucane, and shot him dead in front of his wife and children. The most recent government inquiry into the attack is damning: ‘a series of positive actions by employees of the State actively furthered and facilitated his murder’. And Finucane was not a one-off. While there was no central policy of collusion stemming from Downing Street, various security policies and use of propaganda to discredit certain targets enabled it.⁸

Colluding with rebel and terrorist groups is the most indirect way a state can kill. Collusion has a clear benefit: it becomes difficult, if not impossible, to prove the hand of the state. No orders are given and neither is collusion a policy which can be uncovered, rather a witting – or even unwitting – consequence of a wider culture or approach to a particular conflict. As ever, though, it comes with significant legal and practical problems: the state’s intelligence services will have little influence or control over the deaths. Violence could spiral; innocent people could end up dead. Whether going direct or indirect, or indeed any point on the scale between them, assassinations, like all covert actions, create tough choices. Leaders cannot have it all. An increase in secrecy creates a decrease in control and impact.

Even if MI6 officers could kill people (and they cannot) – it hardly amounts to a licence given the amount of ministerial oversight required. And besides, MI6 does not have paramilitary capabilities. If ministers did authorize lethal force, intelligence officers would rely on their close relationship with special forces to carry it out.

In summer 2021, The Sun breathlessly hyped the existence of ‘the real 007s… a unit of real-life James Bonds so secret [the] government won’t admit they exist’: the elite E Squadron, which works at the disposal of MI6 and the director of special forces. It is a very real unit, staffed by brave personnel and doing important, dangerous work. The journalist quoted one operative: ‘Is it a licence to kill? It is certainly not carte blanche. But the nature of soldiering means it’s sometimes necessary to take life. Everyone is trained in deadly force.’¹⁰ The article – and plenty of others like it – is a classic example of constructing specialness, peddling myth and misperceptions about the licence to kill. Suggestive photographs of Princess Diana (whom MI6 did not assassinate) and interviews with special forces ‘legends’ describing their ‘mad ops’ provided plenty of innuendo. It is little wonder that myths emerge.

Neither do MI6 agents have a licence to kill. Although officers and agents are often wrongly conflated, intelligence agents are those individuals – usually foreign nationals – recruited by officers to spy for the UK. Under certain circumstances, agents can commit crimes in order to prot...