A majority of Americans believe artificial intelligence is barreling ahead faster than it should, and most want the US government to force the technology under human control before the midterm elections, according to a new poll from The Associated Press and the University of Chicago's NORC Center for Public Affairs Research.
The survey, conducted September 24-28 and published this week, found that 64% of US adults think AI technology in the United States is developing too quickly, while just 8% say it is moving too slowly. A quarter of respondents described the current pace as about right. The findings land at a moment when Washington is already scrambling to respond to a string of AI agent incidents, including the false police tip submitted by an Anthropic model and the White House's new AI incident reporting mandate — developments we have been following in our ongoing AI policy coverage.
Worry Crosses Party Lines
The anxiety is not confined to one side of the political spectrum. Among Democrats, 76% say AI is developing too fast; among Republicans, the figure is 54%. Independents and partisans alike, in other words, overwhelmingly share the sense that the technology is outrunning its governance — a rare point of consensus in a polarized electorate.
That consensus extends to what respondents want the government to do about it. Eight in 10 Americans say the US government should prioritize protecting US workers and keeping AI under human control, according to the poll. Similar shares of Democrats and Republicans rate keeping AI under human control, protecting workers, and promoting technological innovation as extremely or very important priorities for the government's approach to AI.
The partisan picture gets murkier beyond those core priorities. Democrats and Republicans diverge more sharply on whether minimizing environmental harm from AI and competing with other countries should be top-tier government objectives — a split that mirrors the broader debates playing out in Congress and on the campaign trail.
Neither Party Has Earned Voters' Confidence on AI
If the poll carries an implicit warning for both parties ahead of the midterms, it is this: voters do not currently trust either of them to manage the technology. Just 31% of adults approve of how President Trump is dealing with AI. Asked which party would do a better job handling AI, 23% trust the Democratic Party more, 21% trust the Republican Party more, and 39% say neither can be trusted to do a good job.
Those numbers arrive as AI policy has moved from the margins to the middle of national politics. Senator Mark Warner introduced the Artificial Intelligence Risk Management and Security Act of 2026 this month, states including California have enacted their own AI laws, and the White House has begun mandating incident reporting after a series of agent failures. Yet the AP-NORC data suggests that whatever officials are doing, a large share of the public remains unconvinced that anyone is actually in the driver's seat.
AI Barely Registers as an Election Tool
There is one notable gap between how much Americans worry about AI and how much they actually use it: despite wall-to-wall coverage of AI's role in politics, the technology is barely present in how voters engage with elections.
Only 21% of adults say they have used or plan to use AI to learn about a candidate or an issue on the ballot. The numbers fall off a cliff from there — 12% for finding information about how to vote, 10% for voter registration, and just 10% for deciding whom to vote for. In an election cycle that pundits have dubbed the first AI election, most Americans are still getting their political information the old-fashioned way.
How the Poll Was Conducted
The nationwide survey was conducted September 24-28, 2026, using the AmeriSpeak Panel, a probability-based panel operated by NORC at the University of Chicago. Interviews were conducted online and by telephone with 2,140 adults, including landline and cellphone respondents. The margin of sampling error is plus or minus 2.9 percentage points.
Methodology matters here: AmeriSpeak is one of the few panels built on probability sampling rather than opt-in respondents, which is why AP-NORC results are treated as a benchmark for measuring American public opinion. The topline finding — that roughly two-thirds of the country thinks AI is moving too fast — is consistent with a string of 2026 surveys from Quinnipiac and Reuters/Ipsos that found deep public unease about the pace and safety of AI development.
For lawmakers heading into November, the poll's message is hard to misread. Voters want innovation, but they want guardrails more — and they are not currently crediting either party with providing them. As the midterm campaigns enter their final stretch, expect AI to feature less as a futuristic promise and more as a test of which party can convince a skeptical public that somebody, somewhere, still has their hand on the controls.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
