Most Americans Want More AI Regulation, USC Survey Finds Before Midterms

Most Americans want the government to do more to regulate artificial intelligence and the data centres that power it, according to a survey released on Monday, November 2, 2026, by the University of Southern California, on the eve of the midterm elections. Overall, 76% of Americans say the government is doing too little to regulate AI, and 78% say the government is doing too little to regulate data centres, a sentiment the researchers say cuts across party lines, levels of AI use, geography and generations.

The findings come from a survey conducted for the new season of the USC podcast Electric Futures: AI, Data Centers, and Trust, which used the nationally representative Understanding America Study to poll 2,980 Americans in October 2026. The survey was published by USC Annenberg's Center for Climate Journalism and Communication, with analysis from the USC Schaeffer Institute for Public Policy and Government Service. Its release, a day before voters decide control of Congress, puts a rarely polled question at the centre of the election-eve debate: not whether Americans use AI, but whether they trust the companies building it and the government overseeing it.

Key Points

• 76% of Americans say the government is doing too little to regulate AI, and 78% say the same about data centres, in a USC survey of 2,980 people polled in October 2026.
• The demand for more regulation crosses party lines: 86% of Democrats and 68% of Republicans say the government does too little on AI, with Independents and others in between at 72%.
• Even daily AI users back more oversight, with 73% saying the government is doing too little to regulate AI and data centres.
• 71% of Americans say they do not trust big tech companies at all to do what is in society's best interest on data centre development, more than the 59% who distrust the national government on the issue.
• The findings were released on November 2, 2026, alongside season 4 of USC's Electric Futures podcast, as AI policy is set to feature heavily in the lame-duck Congress after the midterms.

What Happened

The survey results were released on Monday, November 2, 2026, by USC Annenberg's Center for Climate Journalism and Communication. The researchers polled 2,980 Americans in October using the Understanding America Study, a nationally representative panel, and asked whether the government is doing too little, about the right amount, or too much to regulate AI and data centres. The headline result was unambiguous: in both cases, roughly three in four Americans chose "too little".

The partisan breakdown is the finding the researchers highlight first. Democrats were the most likely to say the government is doing too little, at 86% for AI and 88% for data centres. A smaller, but still clear, majority of Republicans agreed, at 68% for AI and 69% for data centres. Independents and others fell in between, at 72% for AI and 73% for data centres. In other words, while Democrats are more enthusiastic about regulation, no major political group in the survey is satisfied with the current level of oversight.

The same pattern holds when the survey is broken down by how often people actually use AI. Daily users were less negative than less frequent users, but 73% of daily users still said the government is doing too little to regulate AI and data centres. The share was 77% among weekly users, 78% among monthly users, 84% among less frequent users and 76% among non-users. Familiarity with the technology, the survey suggests, softens scepticism only slightly; it does not remove it.

Age and geography tell a similar story. Among Gen Z and Millennials, 74% think the government is doing too little to regulate AI and 75% say the same of data centres. Older generations were slightly more likely to want more regulation, at 79% for AI and 77% for data centres. By region, 77% of Americans in the Northeast say the government is doing too little to regulate AI, compared with 75% in the Midwest, 73% in the South and 80% in the West, with similar numbers for data centre regulation. There is no region, generation or party in the survey where a majority is content with the status quo.

Background: Why AI and Data Centres Are a Political Issue Now

The survey lands in the middle of a fast-moving argument about where AI infrastructure should be built, who pays for it, and who decides. Data centres, the vast buildings full of servers that train and run AI systems, have become a local political issue across the United States, as communities weigh jobs and investment against electricity demand, water use and the pace of construction. At the national level, the same argument has become a question about rules: whether the most powerful AI systems should face mandatory safeguards, and whether local residents should have a formal say in where data centres go.

That debate is about to move into Congress. The lame-duck session opens on November 9, after Tuesday's midterms, and technology companies have already begun positioning for it. OpenAI, for example, has prepared a post-midterm policy push that would rebrand its data centres as "AI compute campuses" and propose so-called speed limits for the most powerful AI systems, alongside community compacts giving localities a say in data centre siting, according to reporting by Axios and TechCrunch. The USC survey does not test any specific proposal of that kind, but it suggests the political ground for such a debate is unusually broad: voters across parties say they want more, not less, government involvement.

The survey also arrives as AI itself becomes an election issue in a different way. Campaigns, candidates and voters are all using AI tools, while regulators, state legislatures and courts are still working out how AI-generated content, automated decisions and large-scale computing should be governed. The USC researchers frame their work as part of that wider conversation about trust: season 4 of Electric Futures, which launched alongside the survey, is devoted to AI, data centres and trust, and will analyse these and other findings with experts.

AI Regulation: The Trust Gap Behind the Numbers

The most striking number in the survey may not be the 76% or the 78%, but the 71%. That is the share of Americans who say they do not trust big tech companies at all to do what is in society's best interest when it comes to developing data centres. Distrust of big tech on this issue is even more common than distrust of the national government, which stands at 59%, according to the survey.

The researchers put that comparison in context. The national government was the least trusted of 25 rated sectors in Gallup's 2025 annual industry ratings, ranking below even the pharmaceutical industry. If Americans trust big tech even less than the national government on data centre development, the survey suggests, then neither of the two institutions most closely associated with AI governance enjoys public confidence on this question. Even most Republicans and daily AI users do not trust big tech companies, the researchers note, although both groups are slightly less negative about big tech and the national government than other Americans.

That trust gap matters because regulation is, in practice, a question of who writes the rules and who enforces them. A public that distrusts both the companies and the government is unlikely to be reassured by industry self-regulation alone, but it may also be sceptical that government action alone will solve the problem. The survey does not resolve that tension. It measures attitudes, not policy preferences in detail, and it does not ask respondents which specific rules they would support, how strict those rules should be, or which level of government should set them. What it does show is a broad, cross-party appetite for the government to do more, whatever form that "more" eventually takes.

Reactions and Quotes

The survey release itself carries no direct interview quotes, but its authors' framing is pointed. Allison Agsten, director of USC Annenberg's Center for Climate Journalism and Communication and a scholar at the USC Schaeffer Institute, and Wändi Bruine de Bruin, Provost Professor of Public Policy, Psychology, and Behavioral Science at USC and director of research with the Center for Civic Society at the USC Schaeffer Institute, present the findings under the heading "Distrust of Big Tech runs across party lines". That framing is the release's central argument: this is not a partisan finding, a generational finding or a finding driven by people who never use AI. It is a finding that holds, with variations in degree, in every group the survey examined.

The researchers also note what did not change the result. Heavy AI use did not turn respondents into opponents of regulation. Living in a particular region did not either. Being young did not either. In each case, the majority in favour of more regulation remained intact, even where it was somewhat smaller. For policymakers reading the survey on the eve of an election, that consistency is arguably the point: there is no obvious constituency, in this data, for leaving AI and data centre regulation exactly where it is.

Why It Matters

The survey matters for three reasons. First, it arrives at a moment when control of Congress is about to change, or not, and when the next Congress will have to decide whether to legislate on AI at all. A finding that three in four Americans want more regulation gives lawmakers who favour action a public-opinion argument they did not have in such clear form before, and it gives opponents of regulation a harder argument to make.

Second, it matters for the data centre debate specifically. Data centres are often approved, or blocked, at the local level, where a single project can dominate a town's politics for months. A national survey showing 78% support for more data centre regulation does not decide any individual planning dispute, but it suggests that local opposition to data centre projects is not an isolated phenomenon. It is part of a national mood.

Third, it matters because it separates use from trust. Americans are using AI more than ever, and the survey's daily users are among the most engaged people in the country. Yet even they, by a margin of almost three to one, say the government is doing too little. That combination — widespread adoption alongside widespread distrust — is the defining tension of the AI era, and it is likely to shape the policy debate long after Tuesday's votes are counted.

What Happens Next

The immediate next step is the election itself. On Tuesday, November 3, 2026, voters will elect all 435 members of the House of Representatives and 35 senators, and the result will determine which party controls each chamber for the final two years of President Donald Trump's second term. The lame-duck Congress then opens on November 9, when AI policy proposals that have been prepared in advance, including OpenAI's planned push, are expected to be put to the new Washington.

Season 4 of USC's Electric Futures podcast, which launched on November 2 alongside the survey, will analyse these findings with experts in the coming weeks. Further surveys in the Understanding America Study series, and other national polls, will test whether the appetite for regulation holds after the election, and whether it translates into support for specific measures. For now, the survey's message is simple and, by the researchers' account, unusually consistent: Americans, across party, age, region and level of AI use, want the government to do more on AI and data centres.

Frequently Asked Questions

What did the USC survey find? 76% of Americans say the government is doing too little to regulate AI, and 78% say the government is doing too little to regulate data centres.

How many people were surveyed, and when? 2,980 Americans were polled in October 2026 using the nationally representative Understanding America Study, for USC's Electric Futures podcast.

Does support for regulation cross party lines? Yes. 86% of Democrats, 68% of Republicans and 72% of Independents and others say the government does too little on AI, with similar figures for data centres.

Do daily AI users oppose regulation? No. 73% of daily AI users say the government is doing too little to regulate AI and data centres, only slightly below the national average.

How much do Americans trust big tech on data centres? 71% say they do not trust big tech companies at all to do what is in society's best interest on data centre development, compared with 59% who distrust the national government on the issue.

Related coverage: NAACP asks court to block federal agents as election monitors on eve of vote, Ohio early voting record: 1.7 million ballots cast before Election Day, and Trump takes California ballot fight to Supreme Court on election eve.