In the summer of 1914, Europe stumbled into a war that almost none of its statesmen had wanted. Each capital believed it was reacting defensively to the mobilisations of others; each set of defensive preparations confirmed everyone else’s worst fears; and within weeks a continent that had been at peace for a generation was tearing itself apart. The First World War is the classic illustration of a disquieting idea at the heart of international relations: that catastrophe can arise not from the ambitions of the wicked but from the precautions of the fearful. The argument of this essay is that the most dangerous actor in world politics is often not the aggressor but the frightened defender – and that the reason is structural, built into the very condition under which states coexist.

That condition is anarchy. The word does not mean chaos; it means simply the absence of a world government – no sovereign above sovereigns, no agency that can be relied upon to protect one state from another. Within a domestic society, a citizen who feels threatened can call the police; between states there is no police to call. Each state is therefore thrown back on self-help: it must, in the last resort, provide for its own security. And here the trouble begins. The measures a state takes to make itself safe – arming, fortifying, forming alliances — necessarily make its neighbours less safe. Those neighbours respond in kind, which leaves the first state less safe than before. The pursuit of security produces insecurity. The historian Herbert Butterfield called this the “tragic” element in human conflict; the theorist John Herz gave it its enduring name in 1950 – the security dilemma.

What makes the dilemma so intractable is a problem of knowledge. A state can observe another’s capabilities but not its intentions. Worse, the capabilities that a status-quo power builds for defence are usually indistinguishable from those an aggressor would build for attack. A wall may be purely defensive, but a tank, a missile, a fleet, an alliance – these serve equally well to deter or to conquer, and no state can be certain which purpose animates its neighbour. Even a genuinely peaceful state cannot credibly prove that it is peaceful, because an aggressor has every incentive to make exactly the same reassuring claims. Uncertainty about intentions is thus not a temporary ignorance to be cleared up with better intelligence; it is a permanent feature of a world of separate sovereignties.

The political scientist Robert Jervis showed that the severity of the dilemma is not fixed but varies with circumstances. Where defence has the advantage over offence – where it is easier to hold ground than to take it – states can be secure without threatening one another, and the spiral is muted. Where offence dominates, and where defensive and offensive weapons cannot be told apart, fear compounds fear, and even small provocations can tip a system into war. Technology, geography and doctrine therefore shape how acute the dilemma becomes. But they never abolish it, because they never solve the underlying problem: the impossibility of seeing into another state’s mind.

This is why the dilemma is properly called a tragedy rather than a crime. Two states, each wanting nothing more than to survive, each reading its own arming as defensive and its rival’s as sinister, can march step by step into a war that leaves both worse off. There is no villain in the story – only the frightened defender, multiplied. The realist John Mearsheimer draws the bleakest lesson: since a state can never be sure of another’s intentions, and since the price of misjudging a rival can be annihilation, the prudent course is always to accumulate power and to assume the worst. On this reading the security dilemma is not a trap to be escaped but the permanent grammar of great-power politics.

Yet the same logic that manufactures fear leaves room, if narrowly, for its opposite. If insecurity is generated by uncertainty about intentions, then anything that reduces that uncertainty can loosen the dilemma’s grip. The theorists Ken Booth and Nicholas Wheeler distinguish the “dilemma of interpretation” – how to read another’s moves – from the “dilemma of response” – how to answer them; and they argue that states can, through costly and unambiguous signals, gradually persuade one another of benign intent. Arms-control agreements that permit inspection, the deliberate adoption of defensive postures, the slow accretion of trust within institutions: these are the instruments by which some clusters of states – Western Europe after 1945, for instance – have built “security communities” in which war between members has become almost unthinkable. The dilemma can, in favourable conditions, be tamed.

Its contemporary reach is not hard to see. A missile-defence system is defensive to the state that builds it and destabilising to the rival whose deterrent it threatens to neutralise. An alliance that expands toward a great power’s borders is, to its members, a shield; to that great power, a spear. A naval build-up in a contested sea is reassurance to one coalition and encirclement to another. In each case both sides can sincerely believe themselves the frightened defender, and each can be moved by that belief to actions that leave the other genuinely less safe. One need not adjudicate any particular quarrel to recognise the pattern beneath them all.

The security dilemma, in the end, teaches a sobering lesson about what is scarce in international politics. It is not, fundamentally, power or territory or wealth that states most lack, but trust – the assurance that a neighbour’s strength will not one day be turned against them. Because that assurance can never be complete among sovereigns who answer to no higher authority, a shadow of fear falls across even the most peaceful relations, and the work of order is never finished. That the shadow can nonetheless be shortened – that frightened defenders can sometimes learn to reassure one another – is the fragile hope on which every lasting peace depends.