Nate Soares, president of the Machine Intelligence Research Institute (MIRI), has not contributed to his retirement for about ten years, because he believes artificial intelligence could wipe out humanity before then. His remarks, made on a podcast, have been picked up by several media outlets, while Elon Musk advised in January not to worry about retirement, for the opposite reason.
Nate Soares spoke on “The Most Interesting Thing in AI,” the podcast hosted by Nicholas Thompson for the American magazine The Atlantic. “I am not saving for my retirement right now,” he explained. “I have a 401(k) from when I worked at Google, but I haven’t contributed to it in ten years.” A 401(k) is an American retirement savings plan. Before joining MIRI, a nonprofit organization based in Berkeley, California, he was a software engineer at Google and Microsoft. He stopped contributing when he began working full-time on AI alignment, meaning research aimed at ensuring artificial intelligence systems pursue goals compatible with human survival.
“More than 50%” risk of not lasting ten years
Asked how much time humanity has left, Nate Soares replied that, in his opinion, there is a “more than 50% chance that we don’t have ten years left.” “I’m not really someone who believes in a job apocalypse because of AI. I’m someone who believes AI is going to kill all humans,” he added, judging the technology “frankly more dangerous than nuclear weapons,” since a bomb “can level a city, but not a planet.”
He does not fear a malicious AI, but a system that has become autonomous, which could consider humans an obstacle if they tried to stop it. Yet he believes nothing is settled: “I’m doing everything I can to prevent that from happening. I don’t think it’s inevitable.” Nate Soares calls on governments to ban the training of general-purpose AIs and to reserve research for specific uses, such as fighting cancer or self-driving cars. He co-authored the book “If Anyone Builds It, Everyone Dies” with Eliezer Yudkowsky, founder of MIRI, which was released in September 2025.
Other researchers cite lower figures
Nate Soares is not the only one quantifying this risk, even though other estimates are significantly lower. Evan Hubinger, who leads alignment research at Anthropic, estimates it at more than 10% in the next ten years. Geoffrey Hinton, the 2024 Nobel Prize in Physics laureate, speaks of 10 to 20% over a few decades, specifying that it is an intuition and not a calculation.
In a survey by the organization AI Impacts, 51% of the researchers polled gave at least a 10% probability to human extinction or a severe loss of control due to AI. Analysts at the RAND Corporation, for their part, estimate that an AI would need to combine four distinct capabilities to threaten humanity, and that such a threat would take a long time to build.
Elon Musk speaks of abundance
Elon Musk arrives at the same conclusion as Nate Soares, but from the opposite direction. In early January 2026, on Peter Diamandis’s “Moonshots” podcast, the Tesla and SpaceX CEO advised not to “worry about putting money aside for retirement in 10 or 20 years,” adding that “it won’t matter.” According to him, AI and robots are set to usher the world into an economy of abundance, even if the transition promises to be difficult. On January 9, 2026, xAI, his artificial intelligence company, had been absorbed by SpaceX.
