Most Americans back AI safety benchmarks even if progress slows, poll finds

Silicon Valley has spent years arguing that AI needs to move fast. The American public would like a word. A Quinnipiac University poll, conducted between September 24 and 27, 2026, found that 86% of US adults support requiring AI companies to meet safety standards, even if doing so slows technological progress. The numbers behind the consensus A full 91% of respondents said US guardrails for artificial intelligence are important. When asked to choose between safety and innovation, 81% picked safety. Some 73% of respondents expressed concern that AI could pose a threat to human survival. The poll found 53% believe AI will do more harm than good in daily life. Trust is the missing ingredient Among the recent survey findings, 74% of people said they have little or no trust in AI companies to regulate themselves. A separate Reuters/Ipsos poll from September 2026 points in the same direction. It found 73% of respondents say AI firms are not doing enough to prevent harm to society. The Quinnipiac results specifically point toward independent safety benchmarks, meaning a third party or government checks the work rather than companies grading their own homework. Not a new mood A 2025 poll ...