Two-thirds of U.S. adults view advanced AI as a moderate existential threat

An online poll of 2,064 U.S. adults carried out between September 13 and 15 revealed that roughly two-thirds perceive advanced artificial intelligence as carrying at least a moderate existential risk. Specifically, 17% said the danger is almost certain, 20% described it as significant, and 26% labeled it moderate. The sentiment crossed political lines, with 70% of Kamala Harris voters and 60% of Donald Trump voters sharing the concern.

When asked about development policy, 48% favored pausing AI work to mitigate risks, while 31% preferred continued advancement for its potential benefits. Among Harris supporters, 58% backed a pause compared with 44% of Trump supporters. Industry figures such as Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman and Elon Musk have publicly called for slower progress, and about half of respondents believe these warnings are genuine, though 26% suspect they are marketing ploys. The poll was commissioned by one outlet and executed by the London-based research firm Public First, with a margin of error of ±2.2 points.