Protect and Survive
In 1980 the British government published a pamphlet called Protect and Survive, advising households on what to do in the event of nuclear attack. It told readers to build an inner refuge from doors and sandbags, to stockpile fourteen days of food and water, to whitewash their windows, and to label and bury the bodies of anyone who died in the shelter. The pamphlet had been prepared some years earlier for distribution only if war appeared imminent, and when its contents became public it was received with a mixture of horror and derision. Critics pointed out, correctly, that its advice would be close to useless against a large-scale attack. The pamphlet became a byword for bureaucratic absurdity, and inspired a small industry of satire.
Such documents repay rereading now, not to laugh at them, though some passages are hard not to laugh at, but because they are unusually clear examples of an institution trying to write for a catastrophe it cannot forecast. The thing I keep returning to in Protect and Survive, and in its American predecessors from the 1950s with their cartoon turtles and schoolchildren under desks, is the gap between the scale of the threat and the scale of the advice. The threat was civilisational. The advice was domestic. Doors off hinges; tins of food; a radio with spare batteries.
The standard reading is that this gap exposes the pamphlets as propaganda — as ways of making nuclear war seem survivable and therefore thinkable. There is something in that. But there is another reading, which I prefer. The domestic advice was absurd only for the worst case. For a range of lesser cases — a limited strike, fallout drifting from a distant target, a breakdown of supply chains, a period of chaos in which the state could not reach you for a fortnight — the advice about water, food, radios and sheltering would have mattered a great deal. The pamphlet failed as a guide to the end of the world. It was a fairly reasonable guide to the bad week before help arrived, and the error of its critics, and perhaps its authors, was to treat those as the same thing.
The same distinction runs through all serious thinking about AI preparedness. There are scenarios — some catastrophic misuse events, or a genuine loss of control over highly capable systems — against which household preparation is roughly as useful as whitewashed windows against a thermonuclear warhead. And there is a much larger set of scenarios, many of them considerably more likely, in which household preparation matters enormously: a lost job, a crashed pension, a drained bank account, a cloned voice on the phone to your father, a week in which the payment systems are down. A sane approach does not pretend the first set does not exist. Nor does it allow the vividness of the first set to distract from the mundane, tractable work of preparing for the second.
from Surviving the AI Apocalypse (2026)
