2026-10-23
AI risk reading list
Where to start if you want to take AI risk seriously. Ordered for a first reader.
If Anyone Builds It, Everyone Dies
Eliezer Yudkowsky, Nate Soares · 2025
Two long-time alignment researchers make the bleakest case: build superintelligence on today's path and humanity likely does not survive. Blunt and short, for readers who want the worst case argued in full.
Human Compatible: Artificial Intelligence and the Problem of Control
Stuart Russell · 2019
The co-author of the standard AI textbook explains why giving machines fixed objectives is itself the danger, and proposes AI that stays uncertain about what humans want.
The Alignment Problem: Machine Learning and Human Values
Brian Christian · 2020
Real cases of machine learning systems learning something other than what their builders intended. No technical background needed; the most readable introduction to alignment.
Superintelligence: Paths, Dangers, Strategies
Nick Bostrom · 2014
The book that brought superintelligence risk into mainstream debate. Dense and philosophical, but most later discussion starts here.
Life 3.0: Being Human in the Age of Artificial Intelligence
Max Tegmark · 2017
A physicist sketches the range of futures AI could bring, from utopia to extinction, and what choices we can still make now.
The Precipice: Existential Risk and the Future of Humanity
Toby Ord · 2020
An Oxford philosopher puts rough odds on each existential risk, from nuclear war to pandemics, and ranks unaligned AI as the largest this century.
The Coming Wave
Mustafa Suleyman, Michael Bhaskar · 2023
A DeepMind co-founder's insider view of why AI and synthetic biology are so hard to contain, and what containment might look like.