Quinnipiac Poll Finds 86% Support AI Safety Standards Even if Development Slows

Americans are willing to accept slower AI development in exchange for independent safety requirements. In a Quinnipiac University poll released September 30, 86% supported requiring AI developers to meet independent safety standards, even if development slows. That preference sits alongside a substantial trust gap: 74% expressed little or no trust in the leaders of AI companies.

The national survey goes beyond asking whether people worry about AI. It asks what they would accept to address that worry, including slower progress and outside standards. Nine percent opposed the independent-standard requirement. Separately, 91% considered U.S. guardrails for AI very or somewhat important, and 81% chose safety over staying at the forefront of innovation.

A slowdown is more popular than a full stop

When respondents chose among four possible development speeds, the largest group favored slowing work on powerful AI systems until safety could be evaluated. A smaller group wanted work stopped for that evaluation. The distinction matters: the combined preference for caution does not mean everyone in that group supports a halt.

• Slow down until safety can be evaluated: 47%.
• Stop until safety can be evaluated: 30%.
• Keep developing at the current pace: 14%.
• Move faster: 5%.

Trust in company leadership was also uneven. Just 3% reported a lot of trust, while 21% reported some. At the other end, 45% said they had none at all, and 29% said not so much. Quinnipiac computer science professor Chetan Jaiswal interpreted the findings as a demand for safeguards that do not depend on taking company leaders at their word.

52%More concerned about humans using AI to cause harm

Asked which concerned them more, 52% chose humans using AI to harm humanity.

33%More concerned about autonomous AI causing harm

In the same question, 33% chose autonomous AI agents doing harm to humanity.

Concerns reached from daily life to human survival. Fifty-three percent expected AI to do more harm than good in their day-to-day lives, compared with 34% expecting more good than harm. That balance was little changed from March. Meanwhile, 73% were very or somewhat concerned that future AI systems could potentially threaten human survival.

Competition and safeguards share supporters

The safety preference does not translate neatly into indifference about international competition. Sixty-nine percent said it was very or somewhat important for the United States to keep up with China in developing AI systems. The question concerned keeping up—not necessarily maintaining a lead.

Quinnipiac found that 65% held both views: keeping up with China matters, and establishing AI guardrails matters. Jaiswal cautioned that this overlap does not show how people would resolve every tradeoff. It does show that supporters of competition and supporters of safeguards are not two separate camps.

Local data centers face stronger opposition

The survey also tested acceptance of AI's physical infrastructure. Seventy-two percent said they would oppose an AI data center in their community, compared with 65% in March. Support stood at 21%, down from 24%, while the share offering no opinion fell from 11% to 7%.

Quinnipiac business professor Tamilla Triantoro said securing local support could become one of the industry's most immediate challenges. These answers measure stated willingness to host a facility; the question asks about an AI data center in general, rather than a named construction proposal with specified local benefits or costs.

The poll surveyed 1,202 U.S. adults from September 24–27, 2026. Its margin of error was plus or minus 3.5 percentage points, including the design effect. Quinnipiac used a probability-based sample, with live interviewers calling landlines and cell phones. The results describe public attitudes toward AI, not measurements of the technology's safety or predictions of its actual effects.