AI Development Speed Poll 2026: What Americans Really Think

The AP-NORC AI development speed poll, released October 8, 2026, found that 64% of U.S. adults say AI technology is developing too quickly, while 27% say the pace is about right and 8% say it is too slow. That single figure is the headline most people reach for, but it becomes far more useful once you see how the AP-NORC survey was conducted and what it asked.

The partisan picture is just as telling. In the same survey, 76% of Democrats and 54% of Republicans said AI was developing too quickly. The poll interviewed 2,140 adults online and by telephone from September 24 to 28, 2026, with a margin of sampling error of plus or minus 2.9 percentage points. Other 2026 polls point in the same general direction, but they often word their questions differently, such as asking whether AI is moving faster than expected or whether development should slow. Checking field dates and question wording before you compare numbers will keep your comparisons accurate.

Key Takeaways

• You can trust the headline that 64% of U.S. adults in the AP-NORC poll (fielded September 24 to 28, 2026) said AI is developing too quickly, a finding that Pew’s 63% and Economist/YouGov’s 65% also support.
• You should check the exact question wording before comparing polls, because the Quinnipiac poll’s 51% measured whether AI was moving faster than expected.
• You can expect both parties to lean the same way on AI speed, since 76% of Democrats and 54% of Republicans in the AP-NORC survey said AI was developing too quickly.
• You should know that 58% of likely voters in the AI Policy Institute’s September 2026 survey said government should already require AI companies to slow development, and 65% said the United States should slow down even if China did not.
• You should only call a change in AI speed concern a trend when two polls ask the same question at similar times and the gap is larger than the margin of sampling error, which is plus or minus 2.9 percentage points in the AP-NORC survey.

How Many Americans Say AI Is Moving Too Fast?

Yes. Most American adults surveyed in 2026 say AI is developing too quickly. The best-known figure is AP-NORC’s 64%, and the other major polls reach a similar majority even when they ask different questions.

Poll	Field dates	Question	Result
AP-NORC	September 24–28, 2026	Is AI developing too quickly, about right, or too slowly?	64% too quickly
Pew Research Center	February 17–23, 2026	Is AI advancing too quickly, about right, or too slowly?	63% too quickly
Economist/YouGov	May 29–June 1, 2026	Is AI advancing too quickly?	65% too quickly
Quinnipiac University	March 2026	Is AI moving faster than expected, as fast as expected, or slower?	51% faster than expected

Pew Research Center’s survey of 5,119 U.S. adults, fielded February 17 to 23, 2026, found 63% saying AI was advancing too quickly. Only 2% said too slowly, 19% said about right, and 16% were unsure. You can read the full findings in Pew’s report on Americans and AI.

A May 29 to June 1, 2026 Economist/YouGov poll put the figure at 65%. Within that poll, 71% of Democrats, 59% of independents, and 66% of Republicans said AI was advancing too quickly.

The Quinnipiac result is smaller because it asks a different question. It measures expectations rather than a judgment about pace.

The AI Policy Institute’s September 2026 survey of 1,047 likely voters, described in its pacing poll, found 71% saying AI was advancing faster than society could keep up with. Verasight, which surveyed 1,500 adults from September 16 to 21, 2026, found 62.6% wanted AI development to slow.

Do Republicans and Democrats Feel the Same Way?

Not exactly, but majorities in both parties say AI is developing too quickly. Democrats are more likely than Republicans to say so in every poll that reports results by party, yet both camps land on the same side of the speed question.

Poll	Democrats	Independents	Republicans
AP-NORC, September 24–28, 2026	76% too quickly	,	54% too quickly
Economist/YouGov, May 29–June 1, 2026	71% too quickly	59% too quickly	66% too quickly

The gap in the AP-NORC survey is 22 percentage points, with Democrats at 76% and Republicans at 54%. The Economist/YouGov poll shows a much narrower 5-point gap, with Republicans at 66% and Democrats at 71%. The two surveys used different field dates and wording, so you should treat the size of the gap as a rough signal rather than a precise measure. Gender also shows a split. In Pew Research Center’s gender breakdown, 68% of women and 58% of men said AI was advancing too quickly. Women are more concerned on this measure, but men still form a clear majority. Where you see agreement is in the baseline. Even the lowest partisan figure, 54% of Republicans in the AP-NORC survey, is a majority, and independents in the Economist/YouGov poll also cross that line at 59%.

Comparing the Latest AI Pace Polls Side by Side

The table below lines up six 2026 polls by field dates, sample, question wording, and headline result. Because the questions differ, read the “Question asked” column before you set two numbers side by side.

Poll and field dates	Sample	Question asked	Main result
AP-NORC, September 24–28, 2026	2,140 U.S. adults	Which best describes the pace of AI development: too quickly, about right, or too slowly?	64% too quickly
Pew Research Center, February 17–23, 2026	5,119 U.S. adults	Which best describes the pace of AI advancement: too quickly, about right, or too slowly?	63% too quickly
Quinnipiac University, March 2026	Not listed here	Is AI moving faster, as fast, or not as fast as expected?	51% faster than expected
Economist/YouGov, May 29–June 1, 2026	Not listed here	Is AI advancing too quickly?	65% too quickly
Verasight, September 16–21, 2026	1,500 adults	Should AI development speed up, continue at its current pace, or slow?	62.6% wanted development to slow
AI Policy Institute, September 2026	1,047 likely voters	Is AI advancing faster than society can keep up with?	71% faster than society could keep up

The Quinnipiac University release asks about expectations rather than judgments of pace, so its 51% answers a question about surprise.

Polls that offer policy choices tell a different story. In Verasight’s survey, 73.4% preferred a hypothetical AI company that develops more slowly and spends more time on safety testing, compared with 11.7% who preferred a faster company with more advanced tools.

The AI Policy Institute offered three options. It found 50% of likely voters chose “pacing,” which it defines as letting AI progress continue while putting mechanisms in place to slow it if progress becomes too fast. Another 33% chose a pause, and 17% chose acceleration. The same survey found 65% said the United States should slow down even if China did not.

Why Does the Wording of an AI Poll Change the Answer?

Each wording measures a different judgment. “Too quickly” asks whether the pace itself is wrong, “faster than expected” asks whether the technology surprised you, and “should slow” asks you to choose a policy. You can hold all three views at once and still answer differently depending on which question you are asked.

The March 2026 poll that used the faster-than-expected framing shows the gap clearly. Its 51% who said AI was moving faster than they expected is a lower share of concern than the 63% to 65% who said “too quickly” in the Pew and Economist/YouGov polls. Being surprised by progress is not the same as believing it is dangerous.

Policy wording shifts answers in another way. The AI Policy Institute’s September 2026 survey of 1,047 likely voters offered three options, and it defined “pacing” as letting AI progress continue while building mechanisms to slow it if progress becomes too fast. Pacing accepts the current speed and prepares a brake for later, which lets people who are uneasy about today’s pace avoid choosing a full stop.

Option offered to likely voters	Share choosing it	What it means
Pacing	50%	Continue progress, with mechanisms to slow it if it becomes too fast
Pause	33%	Stop development for a period
Acceleration	17%	Speed up development

A yes-or-no question about speed cannot capture that middle position. When you compare a “too quickly” figure with a pacing figure, you are comparing a judgment about pace with a preference for a specific policy design.

Should AI Development Slow Down, and Who Decides?

Yes, a majority of likely voters say the government should act. In the AI Policy Institute’s September 2026 survey, 58% of likely voters said the point had already arrived for government to require AI companies to slow development. That is a stronger position than saying the pace feels too fast, because it asks for a rule rather than a personal preference.

Verasight’s survey of 1,500 U.S. adults, fielded September 16 to 21, 2026, points in the same direction. It found 62.6% wanted AI development to slow, 18.6% wanted it to continue at about its current pace, and 4.7% wanted it to speed up. Those figures describe the pace respondents want for development.

The hardest question is whether the United States should slow down if its rivals do not. The AI Policy Institute found that 65% of likely voters said the United States should slow down even if China did not. The table below puts the policy findings side by side.

Question	Share agreeing
Government should already require AI companies to slow	58% (AI Policy Institute, likely voters)
United States should slow even if China did not	65% (AI Policy Institute, likely voters)
Prefer a slower company that spends more time on safety testing	73.4% (Verasight, adults)
Prefer a faster company with more advanced tools	11.7% (Verasight, adults)

The tension is real. Competition with China is usually framed as a reason to keep moving, because falling behind could cost economic and strategic advantages. Yet most likely voters in the AI Policy Institute survey still preferred caution even at the risk of being overtaken. Many people appear to hold both views at once, treating the race as a serious concern while also worrying that speed without safeguards is the bigger danger.

Because the two polls surveyed different groups, with one sampling likely voters and the other adults, read their results as complementary signals rather than a single measurement.

What Worries People About AI Jobs, Data Centers and Control?

Mostly, the worries are about jobs, the data centers that power AI, and whether people keep real control over the technology. The speed question carries all three. When someone says AI is moving too quickly, they often mean they fear losing work, having little say in how it spreads, or being unable to undo changes once they arrive.

The polls covered here do not ask about job worries, but the speed majorities are broad enough that job anxiety is likely part of the picture for many respondents.

Data center opposition is harder to measure. None of the polls reviewed here asks about local facilities, so the best you can do is understand the trade-off residents face. Construction work, tax revenue, and promised community payments can soften opposition, but whether they do depends on who receives the money and how long it lasts.

Control is where the parties most clearly meet. In the AI Policy Institute’s September 2026 survey, 58% of likely voters said government should already require AI companies to slow development, and 65% said the United States should slow down even if China did not. Verasight’s survey found 73.4% preferred a slower company that spends more time on safety testing. These answers ask for oversight, not a particular party’s agenda.

Trust is the gap. None of the reviewed polls asks which party or institution people trust more to handle AI. The policy results do show majority support for government action: 58% of likely voters said government should already require AI companies to slow development.

Long-term pessimism links these concerns. The polls here do not measure how many people expect AI to cause lasting harm, but the logic is straightforward: if the effects are hard to reverse, the speed at which they arrive matters far more.

How Should You Read Margins of Error and Poll Timing?

You can treat a margin of error as the range where the true figure most likely falls, and you can call a change meaningful only when two polls ask the same question at similar times. The AP-NORC survey reports a margin of sampling error of plus or minus 2.9 percentage points, so its 64% figure for “too quickly” plausibly sits anywhere from about 61% to 67% in the full adult population.

Sample size shrinks that range. AP-NORC interviewed 2,140 adults online and by telephone through NORC’s probability-based AmeriSpeak Panel, while Pew’s American Trends Panel drew 5,119 adults. A bigger sample tightens the estimate, but it cannot repair a question that is worded differently from the one you are comparing.

Field dates matter just as much. Pew’s fieldwork ran February 17 to 23, AP-NORC’s ran September 24 to 28, and the Economist/YouGov poll ran May 29 to June 1. Public opinion can shift across months as news about AI breaks, so a difference between those windows may reflect real movement or simply a different news cycle.

Consider the Pew and AP-NORC results. Pew’s 63% in February and AP-NORC’s 64% in September differ by one point, which is far inside the margin of error, and the wording is close but not identical.

Before you call something a trend, check three things: the question wording, the field dates, and whether the gap is larger than the margin of error. Concern about AI speed is broad across the 2026 polls, but the change over time depends on asking the same question again.

Putting These Poll Results to Work

Start by deciding which question you need answered. If you want to know how people judge the pace of AI, the “too quickly” wording is the closest match, while the policy-choice questions show what people would accept in practice. Pick the question that fits your decision and stay with it.

Before you cite any number, build a short comparison sheet. For each poll, write down the pollster, the field dates, the sample size, the exact wording, and the response options. If two results differ by less than the margin of error, describe them as close rather than as a change over time.

Next, open the full release for the poll you plan to quote. Look for the question order and the party or age breakdowns, because the questions asked before an AI item can shape how people answer it. Note whether the AI question came first, after other technology questions, or after a policy question.

Finally, watch for the same question in future polls. A single survey can tell you where opinion stands, but only repeated questions with matching wording can show whether concern about AI speed is rising, falling, or holding steady.