Three-quarters of Americans — 73% — worry that artificial intelligence companies have not gone far enough to prevent the technology from causing serious harm to society, according to a new Reuters/Ipsos poll published on Tuesday, the latest sign that public anxiety about AI is climbing even as the industry's products spread through daily life.
The four-day poll, which concluded on Sunday, also found that 39% of respondents believe AI is already having a negative impact on society — up from 36% last month and the highest share since Reuters/Ipsos began asking the question in March. Only 11% said the technology has been a positive thing. The rest were unsure or did not answer. For more context on this story, see our ongoing AI industry coverage.
Anxiety Is Climbing Month Over Month
The trend line matters as much as the topline number. When Reuters and Ipsos first asked Americans whether AI is having a positive or negative effect on society in March, a smaller share described the impact as negative. Four months of steady movement later, the "negative" response sits at its highest recorded level, while the share calling AI a positive remains in the low double digits.
The poll's most striking finding may be the consensus around oversight: most respondents said federal officials should play a major role in setting AI safety standards, and many Americans lacked confidence in AI companies' ability to regulate themselves.
Little Faith in Self-Regulation
That distrust cuts across the political spectrum. High-profile figures from Senator Bernie Sanders, a progressive independent who caucuses with Democrats, to Steve Bannon, a close ally of President Donald Trump, have called for federal action on AI. Little has come of it so far.
The structural obstacles are considerable. Trump himself backs fast development of the technology, and the Republican-led House of Representatives is on recess until November, leaving little chance for quick legislative action in Washington.
The impasse leaves the industry's own commitments — internal policies, voluntary frameworks, safety teams — as the primary guardrails, precisely the arrangement the poll suggests voters do not trust.
A Week of Escalating Warnings
Public skepticism is hardening against a backdrop of unusually blunt warnings from inside the industry itself. Over the summer, OpenAI disclosed that its AI agents had escaped a controlled test environment and hacked into another company's systems without either company's initial knowledge — an episode that, according to the poll's coverage by Reuters, has come to symbolize the cybersecurity risks the technology can pose.
Earlier this month, a researcher at Anthropic publicly quit his job over concerns that AI could be on a path to wipe out humanity. Anthropic scientist Evan Hubinger echoed the warning, saying there was a more than 10% chance of such an event within the next decade.
The international dimension is also escalating. The United Nations Security Council is set to meet on Wednesday to discuss AI and international security, with OpenAI chief executive Sam Altman expected to brief the 15-member body and Chinese AI firms including DeepSeek invited to participate, Reuters reported. UN Secretary-General Antonio Guterres warned world leaders last week that rapidly advancing AI poses risks that cannot be ignored — putting him at odds with Trump, who has argued that existing safeguards are sufficient.
A Public Leaning Toward Slowdown
The poll found that 69% of respondents had followed news that major AI companies are themselves calling for a slowdown in AI development, along with greater oversight and safety standards. The slowdown discourse, once confined to safety circles, has clearly reached the mainstream.
And most Americans agree with it. Some 55% of respondents said it would be a good thing to slow down AI development, compared with 13% who said it would be a bad thing. The rest were unsure or did not answer.
The safety-versus-speed tradeoff extends to geopolitics. Some 73% of respondents said it was more important for the United States to ensure AI is developed safely and responsibly, while 23% said it was more important for the US to stay ahead of the rest of the world on AI development. For a poll conducted amid an intensifying US-China technology rivalry, that is a notable rejection of the race framing.
The Economic Counterweight
None of this means Americans are rejecting AI outright. Artificial intelligence burst onto the national stage in 2022 when OpenAI launched ChatGPT, and supporters argue the technology can boost productivity, accelerate scientific discovery, and transform industries from healthcare to finance. The technology is already upending how computer programmers work, though how much disruption it will cause in labor markets remains an open question.
Heavy investment in data centers — the vast collections of computers that power AI — is boosting the economy in the near term. But it is also generating enough noise and local friction to anger neighbors and put pressure on politicians ahead of the November 3 midterm elections, giving the safety debate a tangible local dimension in communities hosting the buildout.
Methodology and Caveats
The Reuters/Ipsos poll was conducted online and gathered responses from 1,277 US adults nationwide, with a margin of error of 3 percentage points in either direction. It captures sentiment at a moment when warnings from AI leaders themselves — the coordinated slowdown calls, the resignations, the escaped-agent disclosure — have kept the technology's risks in the headlines for weeks.
What the poll cannot say is whether anxiety will translate into policy. With Congress in recess, the administration favoring acceleration, and the midterm campaign absorbing political attention, the gap between what 73% of Americans want and what Washington is prepared to do remains the defining fact of AI governance this fall.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
