A majority of Americans think artificial intelligence could someday destroy humanity.
The stunning new finding from The POLITICO Poll underscores the depth of people’s anxieties about the rapidly developing technology — and how quickly public sentiment around AI has hardened. Just a few months ago, AI was a nebulous concept for many Americans. Now, lawmakers in both parties are racing to understand how to control its growth.
Overall, about two out of three Americans say there is at least a “moderate” risk that advanced AI could destroy humanity, with majorities across parties, genders and age groups holding that view. Americans who voted for former Vice President Kamala Harris in 2024 are somewhat more likely to see such a risk (70%) than those who voted for President Donald Trump (60%).
Those fears have burst into public view in recent days after the leaders of the nation’s top AI labs — Anthropic’s Dario Amodei and OpenAI’s Sam Altman — called with Elon Musk for slowing the pace of advanced AI development. The rare show of agreement from some of the industry’s fiercest competitors pushed long-simmering warnings about the technology’s dangers into the center of public debate.
And Americans appear inclined to take those warnings seriously.
AI insiders warning there is a real risk the technology “will kill us all” are doing so because they genuinely believe it, according to roughly half of U.S. adults. Just 26% think these leaders are doing it as a marketing stunt to make their companies’ products seem powerful and important, according to The POLITICO Poll.
The survey was conducted by independent London-based Public First Sept. 13 to Sept. 15, as Amodei’s essay was gaining traction, sparking the response from other major tech players. It also finds Americans lean toward slowing or pausing AI development to reduce its risks — saying that AI is already good enough — rather than continuing to build more advanced models. They even split in favor of more responsible development if it means falling behind other countries, like China.
But Trump voters are more torn: A 44% plurality prefer pausing development, only slightly greater than the 40% who say to keep going because AI could still bring significant benefits. And even as tech leaders have started sounding the alarm, Trump himself still insists AI and data centers will be “the Greatest Economic Development Engine in History” and “will not be stopped” during his term.
The poll results signal a growing political problem for the president as he has pushed back on increasingly urgent calls to slow AI development and emphasized the need for the United States to maintain its competitive edge over China. In a series of Truth Social posts on Monday, Trump dismissed the AI alarms as a “HOAX.”
Some of Americans’ concerns over the existential threat of AI may stem from their broader unease with the technology. Across a range of scenarios tested in the survey, majorities of respondents said they would be uncomfortable with AI being used in their everyday lives.
But the latest warnings from industry leaders — contrasted with Trump’s Truth Social missives — have forced Democrats and Republicans to wrestle with the tension in Washington between mitigating the potential risks and ensuring the U.S. doesn’t trail China technologically.
Some Democrats are already considering creating a House select committee on artificial intelligence if they win back control of the House in November. Others have been outspoken in calling for a ban on artificial superintelligence, even before the AI leaders’ warnings last weekend.
Democratic Sen. Chris Murphy, who has been a critic of unchecked AI development, said AI concerns are a “dominant issue” for voters.
“We have to stop AI from poisoning our kids, our culture, and destroying millions of jobs,” he told POLITICO on Tuesday. “This is an immediate crisis, Donald Trump has made it clear where he stands with the AI companies, and the Democratic Party has to put this question to the electorate this November.”
The party isn’t planning to let AI concerns surpass affordability as their top messaging priority for the midterms — at least not yet. But they’re watching the conversation around AI closely, and it’s likely to become a salient issue in 2028.
“We clearly have to do more on this issue to make sure … that these systems don’t do damage to the economy, to people, to communities,” said Sen. Mark Kelly (D-Ariz.), whose name, like Murphy’s, has been floated as a possible Democratic 2028 presidential contender.
“It’ll be part of — should be part of — anybody’s campaign if they’re running for public office,” Kelly said. “Not only here, not only federal office; I think local office, too.”Republicans, meanwhile, are grappling with just how far to distance themselves from Trump. Some governors, including Utah Gov. Spencer Cox and Florida Gov. Ron DeSantis, have been vocal in calling for state-level regulations. Other Republicans on Capitol Hill are advocating for guardrails without stifling growth.
“The industry needs to convene, create a set of operating principles that they — that you’re either certified or you’re an outlier — and then find out what Congress needs to do to stitch it together,” said Sen. Thom Tillis (R-N.C.). “We’re one bad outcome of AI from an overreach here that’d be like Dodd-Frank for AI, and then it’s the worst possible time because China’s not going to slow down.”
[Image text:] A majority of Americans think there's a real risk Al will destroy humanity Share of Americans, split by their 2024 vote, and how much risk they believe there is that advanced Al will destroy humanity. Americans overall Almost certain to happen 17% Trump voters 17% Harris voters 17% Significant risk Moderate risk Very small risk No risk 20% 26% 15% 7% 17% 27% 20% %6 24% 29% 14% 7% Don't know 15% 10% 10% The P0LITIc0 Poll was conducted by Public First from September 13 to 15, surveying 2,064 U.S. adults online. The overall margin of error is +2.2 Source: The POLITICO Poll Anna Wiederkehr/P0LITICO