Only 11% of Americans believe AI is having a positive impact on society. That number is not a typo. According to a Reuters/Ipsos poll of 1,277 U.S. adults published September 22, 2026, nearly three out of four respondents said AI companies have not done enough to prevent serious harm. The poll was conducted online, concluded September 20, and carries a margin of error of plus or minus three percentage points.
If you use an AI assistant, a productivity copilot, or an AI-powered search tool, you are already inside this story. The companies behind those products are the ones the public no longer trusts to govern themselves.
Negative Sentiment: The Numbers Behind the Distrust
Public skepticism toward AI has reached its highest point since Reuters and Ipsos began tracking the question in March 2026, and the shift is not subtle.
What the Reuters/Ipsos poll found:
- 73% of Americans say AI firms have not done enough to prevent serious societal harm.
- 39% say AI is having a negative impact on society, up from 36% the prior month and the highest figure in the tracking series.
- Only 11% describe AI’s societal impact as positive.
- 55% say slowing AI development would be a good thing; 13% say it would be a bad thing.
- 73% say safe and responsible development matters more than keeping the U.S. ahead of other countries; 23% prioritize competitiveness.
Most respondents also said federal officials should play a major role in setting AI safety standards, rather than leaving companies to self-regulate.
Safety Over Speed: What the Public Is Actually Asking For
The speed-versus-safety debate looks very different depending on whether you are inside the industry or outside it.
A Reuters/Ipsos poll found that 73% of respondents placed greater importance on safe and responsible AI development than on maintaining the U.S. lead over other countries.
The industry’s standard argument is that slowing down hands the advantage to rivals. The public’s response, at least by these numbers, is essentially: so what. That gap between what companies are prioritizing and what the public is asking for is the central problem here, and it falls to the industry to close it.
What Needs to Change Before Trust Runs Out
The public is asking for federal involvement, but the poll does not resolve who should act or exactly how.
Respondents did not specify which agency, legal authority, or enforcement mechanism they prefer. Those details remain to be determined. What is not ambiguous is who absorbs the cost if distrust keeps compounding.
Consumers may pull back from AI tools. Employers can face internal resistance when deploying AI systems. Investors are increasingly asking harder questions about governance and accountability. The situation echoes the early days of the social-media privacy reckoning, when industry dismissed public concern right up until congressional hearings forced the conversation.
It is worth noting that the poll measures perception, not the actual effectiveness of any company’s safety program. Some developers maintain serious child safety research operations, and the data does not quantify those efforts. But public perception has a way of becoming its own market force, regardless of what is happening inside the lab.
Meaningful third-party audits and clearer disclosure when AI systems make consequential decisions are a credible start. A federal standard would give skeptical Americans something concrete to evaluate, rather than requiring them to take safety on faith from the very companies whose self-regulation they already distrust.
