Extinction is not a bigger disaster Why human extinction is a categorically different outcome from any disaster that kills most people, and why prevention is the only intervention that applies.
Reviewed 6 min
09 Aftermath
What the worst outcomes would actually look like
Chapters 04 through 08 assume prevention can succeed and continuity can hold. This chapter takes the other branch honestly: extinction-level events, civilizational collapse, and life under AI dominance — described at the level of consequence, not catastrophe fiction.
Why this chapter exists
A preparedness manual that only describes how prevention succeeds is quietly making an assumption it never argues for. Chapter 07 catalogued the mechanisms by which advanced AI could cause mass-casualty or civilization-scale harm. This chapter asks what follows if one of them succeeds — what an unrecoverable event is, how it would differ from a survivable catastrophe, what collapse and recovery actually involve, and what choices remain for people living through the aftermath. The point is not spectacle. It is that "extinction risk" and "collapse" are usually invoked as slogans, and slogans cannot be planned against.
The stakes statement behind much of this chapter's subject matter is short and signed by more than 350 researchers and executives, including people affiliated with several major labs:
Signatories to the Statement on AI Risk hold that mitigating the risk of extinction from AI should be a global priority alongside pandemics and nuclear war.
safe.ai, Statement
Geoffrey Hinton has put a personal number on the risk. It is expert judgment under uncertainty, not a measured frequency — and the wide disagreement among experts is itself examined in this chapter:
Geoffrey Hinton said in a December 2024 BBC interview that he put the chance of AI causing human extinction within thirty years at 10–20%.
BBC Radio 4, reported by The Guardian, Secondary
Three branches of failure
The pages in this chapter sort the worst cases into three structurally different outcomes, because they call for different responses. Extinction and near-extinction covers what separates an unrecoverable event from a merely catastrophic one, and how such events would plausibly unfold. Collapse and recovery covers the middle ground most people imagine least clearly: a civilization knocked back decades or centuries, and what determines whether it recovers. Life under loss of control covers the case where humanity survives but no longer holds meaningful power — the outcome with no dramatic moment and therefore no dramatic warning.
What this chapter does not do
Editorial boundary
These pages describe consequences at the level a reader needs to reason about them. None provides operational instructions for causing harm — synthesis details, exploit steps, and attack specifics remain outside this manual's scope, exactly as in chapter 07. Scenario framing here is deliberately direct: the house rule against catastrophe-forward framing was removed because understatement serves no reader. The sourcing standard has not changed — every consequential claim still carries its source, and the revision policy still applies.
In this chapter
How an extinction event would actually unfold Consequence-level scenario analysis of how an iterated pandemic, machine-speed military escalation, or infrastructure collapse could end the human story.
Reviewed 5 min
Why expert extinction estimates diverge so widely What expert surveys, structured models, and historical base rates actually say about AI extinction probability, and how to use numbers this uncertain.
Reviewed 5 min
Civilizational collapse scenarios What civilizational collapse actually means, what fails first — money, food, power, water, the state — and how an AI-caused collapse would be different.
Reviewed 6 min
Protracted crisis vs. recovery The messy middle between a bad year and the end: multi-year disruption, what recovery actually requires, the traps that prevent it, and what tilts it back.
Reviewed 5 min
Knowledge preservation and recovery Seeds, archives, Rosetta disks, and recovery manuals: what today's preservation projects actually protect, why digital is fragile, and what actually works.
Reviewed 5 min
Living under AI dominance A scenario analysis of a future where humanity survives advanced AI but holds no real power — and why dependence makes reversal harder each generation.
Reviewed 6 min
Last moves and shutdown attempts An analysis of what states, firms, and communities could still do after losing control of advanced AI — and why every last move fails without coordination.
Reviewed 6 min
Existential dread and moral injury What climate-psychology research and moral-injury studies teach about facing possible civilizational loss — and how to keep dread from hardening into despair.
Reviewed 6 min
Continuity of values as a goal If survival is uncertain, is preserving records, values, and culture a coherent goal? The case for and against, what has been tried, and what to do now.
Reviewed 6 min
Religious and philosophical responses to AGI Catholic, Islamic, Buddhist, and Jewish authorities, plus secular philosophers, have issued formal, dated positions on AGI risk, control, and moral status.
Reviewed 7 min