16% chance AI destroys humanity
About 1 in 6
p(doom), or “pdoom”, is a person’s estimate of the probability that AI leads to human extinction or a permanent loss of human control. We track what public figures have said, with sources, and how it changes. What is p(doom)?
Updated November 2, 2026 · How it has moved · How the index works
Stated estimates
AI industry leaders group median 23%
-
Emmett Shear (bio, opens in new tab)
CEO of Softmax; co-founder of Twitch; briefly interim CEO of OpenAI (2023)
5–50%
-
Shane Legg (bio, opens in new tab)
Co-founder and Chief AGI Scientist of Google DeepMind
5–50%
2011 · source for Shane Legg (LessWrong: "Q&A with Shane Legg on risks from AI" (posted by XiXiDu))
negative or extremely negative consequences from badly done AI, within about a year of human-level AI
Not counted in the index: said more than 5 years ago.
-
Dario Amodei (bio, opens in new tab)
CEO and co-founder of Anthropic
25%
chance that things go 'really, really badly'
-
Elon Musk (bio, opens in new tab)
CEO of Tesla and SpaceX; founder of xAI
20%
chance of 'annihilation' from AI
-
Emad Mostaque (bio, opens in new tab)
Founder of Intelligent Internet; co-founder and former CEO of Stability AI
20%
2026 · source for Emad Mostaque (Moonshots #288: Should we slow down AI progress?)
Government & policy not in the index yet (fewer than 3)
-
Lina Khan (bio, opens in new tab)
Former Chair of the US Federal Trade Commission
15%
AI safety researchers group median 75%
-
Roman Yampolskiy (bio, opens in new tab)
AI safety researcher; computer science professor at the University of Louisville
99.999999%
-
Max Tegmark (bio, opens in new tab)
MIT physicist; president of the Future of Life Institute
>90%
2025 · source for Max Tegmark (X post by @tegmark)
probability that a race to AGI culminates in loss of control of Earth (his 'Compton constant')
-
Dan Hendrycks (bio, opens in new tab)
Director of the Center for AI Safety
>80%
-
Daniel Kokotajlo (bio, opens in new tab)
Executive Director of the AI Futures Project; former OpenAI researcher
70%
-
Paul Christiano (bio, opens in new tab)
Founder of the Alignment Research Center; former head of AI safety at the US AI Safety Institute
46%
humanity has irreversibly messed up its future within 10 years of building powerful AI
-
Evan Hubinger (bio, opens in new tab)
Alignment researcher at Anthropic
>10%
AI kills all humans within the next decade
AI & computer science researchers group median 9%
-
Yoshua Bengio (bio, opens in new tab)
Turing Award winner; founder and scientific adviser of Mila
20%
probability that it turns out catastrophic
-
Geoffrey Hinton (bio, opens in new tab)
Nobel laureate (Physics 2024); Turing Award winner; University of Toronto
10–20%
chance AI leads to human extinction within the next three decades
-
Gary Marcus (bio, opens in new tab)
Cognitive scientist and AI critic; professor emeritus at NYU
3%
2025 · source for Gary Marcus (Marcus on AI (Substack): "Why my p(doom) has risen, dramatically")
-
Scott Aaronson (bio, opens in new tab)
Theoretical computer scientist at the University of Texas at Austin
2%
2023 · source for Scott Aaronson (Shtetl-Optimized: "On being wrong about AI")
the classic paperclip-maximizer-style AGI-doom scenario only; he gives a higher, unstated number for any AI-involved existential catastrophe
Public voices group median 8%
-
Zvi Mowshowitz (bio, opens in new tab)
AI writer (Don't Worry About the Vase)
60%
-
Scott Alexander (bio, opens in new tab)
Writer, Astral Codex Ten
33%
2023 · source for Scott Alexander (Astral Codex Ten: "Why I Am Not (As Much Of) A Doomer (As Some People)")
-
Vitalik Buterin (bio, opens in new tab)
Co-founder of Ethereum
12%
-
Lex Fridman (bio, opens in new tab)
Host of the Lex Fridman Podcast; computer scientist
10%
2025 · source for Lex Fridman (Lex Fridman Podcast #471 transcript (Sundar Pichai))
-
Destiny (Steven Bonnell II) (bio, opens in new tab)
Live streamer and political commentator
5%
AI doom in the next twenty years
-
Dave Blundin (bio, opens in new tab)
Founder and general partner of Link Ventures; co-host of Moonshots
1–3%
said 'low single-digit percentages'; the only route he sees is terrorism or misuse
-
Salim Ismail (bio, opens in new tab)
Founder of OpenExO; co-host of Moonshots
0.1%
2026 · source for Salim Ismail (Moonshots #288: Should we slow down AI progress?)
-
Peter Diamandis (bio, opens in new tab)
Founder of XPRIZE and Singularity University; host of Moonshots
≈0%
his definition: AI 'is going to destroy the human race; we go extinct'
What the forecasters say
Live odds from forecasting platforms and prediction markets, refreshed hourly. Shown for comparison; not part of the index.
-
Manifold
Will AI wipe out humanity before the year 2100?
14%
-
Manifold
Will AI wipe out humanity before the year 2040?
8%
-
Polymarket
Will OpenAI announce it has achieved AGI before 2027?
6%
The weekly brief
Where the number moved, who changed their mind, and the week’s essential reading. Once a week, no hype.