The poll also found 86 percent support independent safety standards for AI companies, even if those rules slow progress, and 91 percent say guardrails for AI systems in the US are important. Quinnipiac surveyed 1,202 US adults between September 24 and 27, with a margin of error of plus or minus 3.5 percentage points.

Other results point in a similar direction: 74 percent have little or no trust in the leaders of AI companies, 73 percent worry AI could eventually threaten humanity's survival, and 52 percent are more worried about humans misusing AI than about autonomous AI misbehaving on its own. "Taken together, these results suggest that the public wants safeguards that do not depend simply on taking company leaders at their word," said Chetan Jaiswal of Quinnipiac University.

The picture is not uniformly anti-AI. A majority, 53 percent, think AI will do more harm than good in daily life, yet 69 percent say it is important for the US to keep pace with China on AI. And 72 percent oppose an AI data centre in their own community, suggesting support for the technology at a national level sits uneasily alongside resistance to its local footprint.