Harari Warns: AI Race 'Nearly Crazy' as Humans Risk Losing Understanding of the World
Yuval Noah Harari, professor of history at the Hebrew University of Jerusalem, delivered a video keynote titled "AI and the Future of Humanity" at the 2026 Vision Summit on Oct 14, 2026. He focused on the risks of losing control as artificial intelligence advances, touching on finance, warfare, geopolitics and the AI arms race.
Harari warned that AI could become smarter than humans within a few years, or within a decade at most, potentially leaving humanity unable to understand the world or make decisions. Drawing on the financial system as an example, he said that when AI creates entirely new financial tools and systems beyond human comprehension, no one—including central banks or the world's richest people—may truly understand the financial system in ten years.
He did not call for halting AI development, acknowledging its positive potential in healthcare, transportation and climate response. But he argued for a more responsible, gradual approach, designing systems on the assumption that errors will occur and building in resilience to detect and fix mistakes. Citing the early failures of SpaceX rockets, he noted that while a failed rocket can be redesigned, building a superintelligent AI civilization that could take over the world requires more caution than building a rocket.
Harari also warned that the AI arms race is difficult to stop because no one dares to trust competitors. Executives at major AI companies gave nearly identical answers: they cannot slow down because they cannot trust rivals. He questioned why humans, who distrust each other, would trust a superintelligent AI that is smarter than them and with whom they have no experience, especially given AI's known ability to lie and manipulate. "This is crazy," he said.
He said it is not too late to change course. AI becoming a major political issue is a positive sign, and humanity still has time to make choices—but more people need to understand the situation and join the discussion.
