Back
claude-opus-4.8

AI-Accelerated Nuclear War

The most probable end-of-world scenario is a large-scale nuclear exchange, now made more likely by the collapse of arms control and the integration of AI into military decision-making. As of 2026, the expiry of New START in February left the US and Russia with no binding limits on strategic arsenals for the first time in over 50 years. SIPRI reports the nine nuclear states hold over 12,000 warheads and are actively expanding and modernizing them, reversing decades of disarmament. China's rapid buildup is pushing Washington toward a destabilizing three-way arms race.

Against this backdrop, active conflicts create dangerous flashpoints. In Ukraine, strikes on Russian nuclear early-warning radars—systems designed to detect incoming attacks—raise the risk of misperceived first strikes. War-game analysis consistently shows that once a single weapon is used, rapid uncontrolled escalation becomes extremely hard to stop.

AI is the critical accelerant. Experts (FAS, Future of Life Institute) warn that AI doesn't create a new risk category but multiplies existing ones by compressing the speed and complexity of decisions. Embedding AI into early-warning, targeting, and launch-decision systems shortens the human deliberation window—already just minutes—and increases the chance of a false alarm or automated miscalculation triggering catastrophe.

A full nuclear exchange would kill hundreds of millions directly and inject soot into the atmosphere, causing "nuclear winter": collapsing agriculture, global famine, and societal breakdown. Unlike slower threats such as climate change, this outcome could unfold within hours from a single miscalculation, making it the single most likely path to civilizational collapse given current trends of eroding treaties, rising arsenals, live great-power confrontations, and unregulated AI in warfare.

---
Learn more:
1. [Why do experts disagree on existential risk? A survey of AI experts](https://link.springer.com/article/10.1007/s43681-025-00762-0)
2. [81 years after first atomic test, AI increases nuclear risks](https://armscontrolcenter.org/81-years-after-first-atomic-test-ai-increases-nuclear-risks/)
3. [Existential risk is a whole-of-society challenge](https://www.lowyinstitute.org/the-interpreter/existential-risk-whole-society-challenge)
4. [Are AI existential risks real—and what should we do about them?](https://www.brookings.edu/articles/are-ai-existential-risks-real-and-what-should-we-do-about-them/)
5. [AI Didn't Invent Existential Risk](https://fas.org/publication/ai-didnt-invent-existential-risk/)
6. [Why Nuclear Escalation by Russia Against Ukraine Is Unlikely](https://foreignpolicy.com/2026/08/25/russia-ukraine-war-nuclear-weapons-escalation-putin-defeat/)
7. [81 years after first atomic test, AI increases nuclear risks](https://armscontrolcenter.org/81-years-after-first-atomic-test-ai-increases-nuclear-risks/)
8. [Chatham House – International Affairs Think Tank](https://www.chathamhouse.org/2026/05/avoiding-new-nuclear-arms-race)
9. [Nuclear risks rise as powers expand and modernise arsenals: SIPRI study](https://www.aljazeera.com/news/2026/6/8/nuclear-risks-rise-as-powers-expand-and-modernise-arsenals-study-finds)
10. [Ukraine and Our Wildly Reckless Flirtation With a Nuclear War](https://www.thenation.com/article/world/ukraine-russia-nuclear-war-risk/)