AI safety leader says he stopped saving for retirement over extinction fears
MIRI president Nate Soares puts the odds of AI wiping out humanity within ten years above 50%. It is one researcher's judgment, but it shapes the safety debate.
The short version
- Soares says he hasn't added to his 401(k) retirement account in ten years.
- He estimates the chance of AI killing all humans within a decade is above 50%.
- His worry is an AI that doesn't care whether people survive, not one that hates them.
- He wants a pause on training general-purpose AI and treats rogue superintelligence as a nuclear-level threat.
- He says disaster is not inevitable.
Why it matters
Stark predictions from people inside the field get quoted in news coverage and policy talks, so they influence how governments and companies talk about AI risk. The number is a personal estimate, not a measurement. Read it as a view in the argument, not as a forecast you can verify.
What Soares said
As technews.tw reported, Nate Soares says he has not put new money into his 401(k) in ten years. A 401(k) is a US employer-linked retirement savings account. He still holds the one he opened while working at Google, but he says his reason is not a change in financial planning. It is that he thinks AI could pose a deadly threat to everyone within the next decade.
He made the comments on the podcast The Most Interesting Thing in AI. He put the chance of AI causing the death of all humans at above 50%.
Who he is and what he argues
Soares left Google in 2014 and joined the Machine Intelligence Research Institute (MIRI), where he has studied how to make advanced AI act as people intend. In 2025 he and MIRI co-founder Eliezer Yudkowsky wrote a book, If Anyone Builds It, Everyone Dies.
Its argument is that if superintelligent AI is built with today's methods, it could end up with goals quite different from what its makers expected, and those goals could threaten human survival.
Indifference, not hatred
Soares stresses that he does not fear an AI that hates people. His concern is one that simply does not care whether humans live. He argues that a powerful enough system could gather resources, build infrastructure that runs without human help, and keep improving itself, while people gradually lose control.
What he wants done
Soares calls for countries to pause the training of general-purpose AI and focus on narrower, safer uses, such as cancer research and safer self-driving cars. He also wants governments to treat runaway superintelligent AI as a threat at least on par with nuclear weapons.
He says he will do everything he can to prevent that outcome, and he does not believe catastrophe is unavoidable.
How to read the claim
The report does not explain how Soares arrived at his figure, so there is no way to check it. A probability like this is a judgment call, not a lab result. It tells you how seriously one prominent safety researcher takes the risk, not what will happen.
What to watch
- Whether Soares or MIRI publish the reasoning behind the above-50% estimate.
- Whether governments take up his call to pause training of general-purpose AI.
Try it
When you see a bold AI prediction, ask three questions: who is saying it, how did they get the number, and what do they want done about it?
Sources and links
- Reportingtechnews.tw: the article we sourced
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by technews.tw · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.