In certain special forces operations, like a commando raid, a small team moves ahead of everyone else. Two, sometimes three soldiers. Their mission is called sentry neutralisation — the clinical term for what is, in practice, the most intimate act of violence in modern warfare. These soldiers must creep through darkness to within arm’s reach of an enemy sentry, gag his mouth with one hand, overpower his body with the other, and with either a garrotte or a dagger between the ribs, take his life. Silently. Restraining a human body while it is withering spasmodically in its final death throes.
There is no distance in this act. No screen, no trigger, no intermediary. Just two human beings — one killing the other. The soldiers trained for this are not monsters. That is precisely the point.
On Christmas Eve, 1914, during World War I, something happened on the Western Front, that high command on both sides tried desperately to suppress. German soldiers began placing candles along their trench parapets. British soldiers watching from 50 yards away heard singing. Within hours, men who had spent months trying to kill each other were standing in No Man’s Land — sharing cigarettes, exchanging food, and by multiple accounts, playing football. High command scrambled to prevent its recurrence.
What this proved was something military planners had always understood and never wanted to acknowledge: Soldiers do not hate each other. The enmity of soldiers is not personal. It is institutional — manufactured by States, sustained by political rhetoric, and insulated by deliberate dehumanisation that turns a fellow human being into an impersonal target. The Indian armed forces — like several others — reiterates this at every firing range in bold letters: “No pity, no regret & no remorse.”
The sentry silencing party experiences this at a primal level. To kill a man with bare hands, a soldier must first overcome every empathetic instinct of his species. Lieutenant Colonel Dave Grossman, in his landmark study On Killing, documented that 75-80% of World War II soldiers could not bring themselves to fire directly at the enemy — even in active combat. The resistance to killing another human being is not weakness. It is the most fundamental human instinct. And it has to be systematically engineered away.
For most of military history, that primary engineering tool was distance.
A bayonet requires the killer to be close enough to feel the resistance of flesh and tearing of cartilage. A rifle allows him to see the target fall but not the face. An aerial bomb makes the target invisible — a coordinate, a grid reference. A drone allows the operator to rain death through a screen, in air conditioning, before driving home for dinner.
Each step along this gradient increases range and reduces moral friction. This is not incidental. Grossman’s research showed kill rates increased dramatically as weapons technology increased distance. The military did not just build longer-range weapons. It systematically built moral distance — because most human beings, left to their own instincts, are not particularly ardent killers. Their conscience gets in the way.
Now consider what autonomous AI weapons represent. Not a rifle with longer range or a drone with heavier payload — but the removal of a human being from the kill chain entirely.
When Pope Leo XIV condemned AI-directed warfare at a Rome university recently the mainstream reaction was to file it under “church versus technology”. That framing misses the point entirely. The Pope was not making a theological argument. He was giving voice to a psychological discomfort that millions of people already experience but have no authority to articulate. Philosopher Robert Sparrow’s seminal 2007 analysis identified what he called the “responsibility gap”. When a human pulls a trigger, there is a chain of accountability: soldier, commander, State. When an algorithm selects a target, that chain dissolves. The algorithm cannot be court-martialled. The programmer did not choose the victim. The commander who deployed the system points to the architecture. The aggressor says: It was not me. It was the system.
This is not a legal technicality. It is the systematic outsourcing of moral accountability — the most consequential element human civilisation has ever attempted to outsource.
Morality is not just a personal attribute. It is a social contract — a shared system of accountabilities that makes society function. When the decision to take a life is delegated to an algorithm, the moral accountability underwriting that contract is delegated with it. And once a society accepts that algorithms can make lethal decisions without human accountability, the path runs in only one direction.
The COMPAS algorithm in the US already generates risk scores that influence prison sentences — making predictions about human behaviour that affect years of a person’s life, with no human judgment required. During the Covid-19 crisis, hospitals in multiple countries used algorithmic triage to allocate ventilators, deciding on the basis of projected outcomes on who had sufficient systemic or monetary value to receive life-sustaining treatment.
These are not hypotheticals. They are the gradient, hurtling towards a world in which algorithmic systems assign value to human lives — not in service of individual dignity, not in response to the grief of a family, but in optimisation of systemic efficiency. Not for an ideology or for a leader, but for their own optimisation function.
The Christmas truce was possible because human soldiers were making human decisions — and somewhere beneath the training, the propaganda, and the fear, they recognised each other as fellow humans and not perennial enemies, the very reason they are allies today.
The sentry silencing party has to physically clasp the man they are killing. They carry that moment for the rest of their lives. That weight — the moral burden of taking a life — is not a flaw in the human character. It is the last guardrail between a violent act and a civilisational catastrophe. By replacing that judgement, and its accompanying morality, with an algorithm, we may speed up the decision loop. But we will also cross a chasm from which there is no viable return.
Raghu Raman is founding CEO, NATGRID, and a former soldier. The views expressed are personal
