The idea that AI moving too fast is a public concern is supported by a new American survey rather than just the warnings of technology executives. An AP-NORC poll released on 8 October 2026 found that 64% of US adults believe artificial intelligence is developing too quickly. Only 8% said it was moving too slowly, while about a quarter considered the pace about right.
The AP-NORC survey also found broad support for keeping artificial intelligence under human control and protecting workers. It does not establish that respondents agree on which laws should be passed, which applications should be restricted or whether progress itself should stop. That distinction matters because anxiety about an unfamiliar technology can produce very different policy preferences.
Why AI moving too fast worries voters
In the survey, 76% of Democrats and 54% of Republicans said AI development was proceeding too quickly. That is a notable gap, but it also means concern was present in both party groups. Respondents from different political backgrounds shared interest in human oversight, workforce protection and innovation, while diverging more on environmental issues and international competition.
These figures describe attitudes at the time of the interviews, not permanent positions. The research was conducted from 24 to 28 September 2026 with 2,140 adults using NORC’s probability-based AmeriSpeak panel. The reported overall margin of sampling error was plus or minus 2.9 percentage points. Survey wording, timing and sampling always matter when judging how firmly to interpret public opinion.
The Associated Press reporting placed the findings in the wider US debate about increasingly capable models and government oversight. It is useful context, but the poll itself is the strongest evidence for its particular numbers.
People want control without agreeing on the controls
Saying that humans should remain in charge sounds straightforward until it must be translated into rules. In a school, it might mean a teacher reviewing machine-generated feedback. In a bank, it might require a meaningful route to challenge an automated decision. In a workplace, it could mean that software flags a problem but a responsible employee retains authority to act.
The poll cannot tell policymakers which implementation will work best. It does show why claims that consumers inevitably welcome every new capability should be tested against evidence. People may enjoy faster services while also wanting assurance that consequential decisions can be explained and reversed.
Questions about public trust in AI decisions often turn on transparency, consent and practical remedies. A system that appears convenient in a demonstration may feel very different when it affects a loan, a job interview or an unexplained rejection. That is a reasonable interpretation of the wider tension, rather than a direct response recorded from every participant.
Jobs remain central to the argument
The survey identified protecting US workers as an important priority for many respondents. That fits with a broader debate over AI and worker displacement, which includes fears of job losses alongside hopes for improved productivity and new kinds of employment. Neither the survey nor any single employer announcement can settle what the total labour-market impact will be.
The immediate choices for businesses are more practical. What tasks are being changed? Who is consulted? Can employees contest a system’s errors? Is training provided when expectations change? These questions are often more useful than a broad prediction that an entire occupation will vanish or remain untouched.
Public attitudes may also be shaped by how benefits are distributed. A tool that makes an organisation more profitable does not automatically persuade staff that their own position is secure. Conversely, employees may welcome assistance when it removes frustrating work while preserving responsibility and opportunities for advancement.
A survey conducted before an election can also reveal priorities without establishing how people will vote. Respondents may rank AI safety highly while still making electoral choices mainly on other subjects. The poll asked about attitudes to technology and leadership, so its numbers should not be turned into predictions about the result of a future contest. Readers should also avoid assuming that an American sample describes opinions in Britain, France or other countries.
The survey is a signal, not a referendum
The poll also explored confidence in political leadership. It found limited trust in either major party to handle AI particularly well, a reminder that anxiety about rapid technical change does not map neatly onto a single political position. As with other findings, these responses should not be treated as forecasts of voting behaviour.
The significance is that the pace of AI development has become an ordinary public issue, not just an argument between engineers. The results tell companies, governments and researchers that explaining what systems can do is not enough. They also have to explain who authorises their actions, who remains accountable and what happens when something goes wrong.
A poll measures sentiment. It cannot prove that any particular system is dangerous, nor can it calculate an ideal speed of innovation. But it provides a useful check against the assumption that public enthusiasm automatically rises as technology becomes more capable.
The central distinction is between wanting a public discussion about AI and agreeing on one particular policy. A useful public debate would test concrete proposals against their costs, their likely effects and the views of the people most directly affected.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, cybersecurity and the social impact of emerging technology. LiveAIWire is an independent, human-led technology publication using AI-assisted research, editorial production and original AI-assisted editorial illustrations under his direction.