Most Americans think artificial intelligence is developing too quickly — and they agree with each other about what government should do about it, across party lines. A new poll from The Associated Press-NORC Center for Public Affairs Research finds 64% of US adults say AI is moving “too fast,” against just 8% who say “too slow.” The survey of 2,140 adults was conducted September 24-28 and published October 8; it has been recirculating through US papers this week, including the Boston Herald’s Monday run and syndicated copies reaching global audiences.
🔍 THE BOTTOM LINE: The public’s verdict is not anti-AI — it is anti-uncontrolled. Majorities in both parties want the government to keep AI under human control and protect workers, while 39% trust neither party to manage the technology at all. An industry whose CEOs preach acceleration is outrunning the one constituency that matters: permission from the people it is deployed on.
What the numbers actually say
The detail beneath the 64% headline is more interesting than the headline. Agreement on the pace-of-development worry splits by party — 76% of Democrats versus 54% of Republicans say “too fast” — and by age, with under-30s more concerned than over-60s, a reversal of most technology-anxiety curves. (NORC’s own summary)
But the policy priorities converge. Roughly 8 in 10 US adults say it is “extremely” or “very” important for government to keep AI under human control, with nearly identical bipartisan shares wanting worker protection prioritised. The partisan split appears on the other questions: Republicans weight economic growth and China competition higher, Democrats weight environmental harm higher.
And the trust numbers are stark: 31% of adults approve of President Trump’s handling of AI, 23% trust the Democratic Party more, 21% the Republican Party more — and 39% say neither party can be trusted to do a good job. For a technology whose deployment is currently being arbitrated by exactly those two institutions, that is a legitimacy gap.
The gap between public preference and deployed policy
The poll ran September 24-28 — the same week the White House asked OpenAI and Anthropic to withhold new frontier models from Britain’s testers, and the administration’s AI czar was given a 120-day window to report on AI risks leaned towards industry self-policing rather than binding rules. What the public tells pollsters it wants — eight in ten for human control — and what the policy apparatus is building — voluntary commitments, review-first frameworks without published findings — sit in different postcodes. This site has tracked the pattern all year: the UK rejected its own kill-switch amendment in September confident in cooperative testing (/uk-ai-kill-switch-rejected-cabinet-office-2026/), and the cooperative testing then turned out to be revocable at one government’s discretion (/white-house-restores-uk-aisi-access-frontier-models-2026/).
The worker-protection consensus has a concrete policy echo too. California’s No Robo Bosses Act — human corroboration required for AI-only firing decisions, signed September 30 — polls exactly in the territory this survey maps: our coverage noted business groups lost an argument most voters, including most Republicans, would probably side against them on. Workplace AI rules are the rare issue where the median voter and the median union official agree.
Why this poll matters more than its genre
Opinion polls about AI arrive weekly and mostly tell you what people already said last year. Two things make this one worth sitting with. First, the composition of the worry: the youngest adults are the most concerned about pace — the cohort assumed to be native adopters — which undercuts the industry’s “the kids are fine with it” defence. Second, the specific shape of the consensus: “under human control” and “protect workers” are not abstraction questions; they map directly onto the agent-transparency and displacement-disclosure rules being written (or not written) right now.
The honest caveat: worried publics habituate. The “too fast” share has been in the sixties across pollsters for two years without producing commensurate federal legislation. Nor is concern a verdict on any specific use — the same respondents use AI tools at work and school. But as a permission structure, the numbers say the industry’s social licence is issued on conditions — control and workers first — that current policy is not meeting. New Zealand readers will recognise the shape: this site has tracked NZ workers as among the most relaxed about AI in the world (/employment-hero-nz-ai-paradox-workers-2026/), which makes the US poll a useful stress-test of assumptions here — NZ’s comfort is not the global distribution; it is the outlier.
As of publication, NORC has published crosstabs at the project page linked above; the margin of sampling error is +/- 2.9 percentage points.
❓ FAQ
What did the poll find? 64% of US adults think AI is developing too fast (76% of Democrats, 54% of Republicans), 27% say the pace is about right, 8% say too slow. About 8 in 10 want government to prioritise keeping AI under human control, with similar bipartisan majorities prioritising worker protection.
Who was surveyed? 2,140 US adults, September 24-28 2026, via NORC’s probability-based AmeriSpeak Panel; margin of error +/- 2.9 percentage points.
Do Americans trust either party on AI? Barely: 23% trust Democrats more, 21% Republicans more, 39% neither — and 67% disapprove of President Trump’s handling of AI.
Is there a New Zealand comparison? NZ workers polls unusually relaxed about AI (Employment Hero, September 2026), but NZ also has no AI-specific employment law on the statute book — the US poll suggests the control-and-workers consensus is where most developed-market publics sit.
🔍 THE BOTTOM LINE
The poll’s real finding is the shape of agreement, not the headline number. A country that cannot agree on almost anything agrees that AI should stay under human control and that workers come first — and 39% think neither party is listening. Whatever the technology’s actual trajectory, the political permission to run it fast is thinner than the industry’s investors like to believe.
📰 Sources
- AP-NORC Center for Public Affairs Research — Most adults think AI is developing too fast (8 October 2026)
- AP (via BastillePost syndication) — Most Americans think AI is developing too fast, a new AP-NORC poll finds
- Boston Herald — Most Americans think artificial intelligence is developing too fast (8 October 2026)
