AI Regulation

Across Democrats, Republicans, and Independents, majorities of Americans want to see more regulation of AI (artificial intelligence). Some states have begun passing AI laws, mostly requiring companies to disclose or document their practices. Some localities have gone further by blocking or delaying AI data-center projects. At the national level, the Trump administration has moved in the opposite direction, pressing for lighter regulation and seeking to preempt state laws.

With the government response mixed, what pathways exist from public opinion to policy?

Concern Across Party Lines

Polls have shown consistent support for AI regulation, across party lines. For example:

People’s reasons for wanting regulation vary widely. Among the concerns cited:

• An AI could escape human control and cause large-scale harm to society.

• AI agents could replace people in jobs at mass-scale.

• AI used in decisions like who gets a loan or who gets into college could be biased based on historical training.

• AI data centers’ demand for electricity can cause prices to increase for everyone.

• AI data centers may harm the environment through their water consumption for cooling and through emissions from fossil-fuel power that often runs them.

Each concern points toward a different policy response. Ask Americans not just whether AI should be regulated but what regulators should actually do, and the cross-partisan majority can fracture: there is no comparable consensus on how to handle job loss or algorithmic bias. But two policy directions have cross-partisan majorities: prioritizing safety over speed, and resisting AI data centers nearby.

Safety Over Speed

As AI becomes increasingly capable, and is more widely used, it also has greater potential for causing harm. AI companies try to design their systems for safety, but any guardrails can potentially be defeated. For example, a sophisticated attacker could trick an AI into doing harm, or an AI could by itself get out of control in a bad way. These are hypotheticals, with unknown probabilities over time.

Given this uncertainty, there have been moments when even some AI advocates have called for a pause to ensure that safety measures and policies can keep up with the rapidly advancing technology. Thus far, no pauses in technology development have occurred.

Nevertheless, cross-partisan majorities said the government should regulate AI to prioritize safety over speed:

Most of the leading AI companies and the Trump administration have been against pauses. A key reason is that the United States is in a race against China to have the most advanced AI models. Having the best models is important to future economic and military success, perhaps decisively so. Pausing would jeopardize the U.S.’s advantage, say those against pausing.

Recently, Anthropic (the company behind the AI assistant Claude) said, “it would be good for the world to have the option to slow or temporarily pause frontier AI development.” But the company emphasized that this would only make sense if the pause could verifiably include all large-scale global players—a potentially intractable requirement.

Meanwhile, the leading AI companies argue that they are not ignoring safety. They publish voluntary safety frameworks and run internal and third-party “red-team” testing meant to surface dangerous capabilities before a new AI model is released. They argue, too, that keeping the frontier in American hands is itself the safest course: if the U.S. slows down, a Chinese lead would set the global norms instead. The Trump administration’s 2025 AI Action Plan leans on this logic, favoring voluntary standards developed with the National Institute of Standards and Technology over binding federal rules. Whether voluntary measures are enough is exactly what the public seems concerned about.

A recent episode twisted this question when the Trump administration sought to end use of Anthropic’s AI at the Pentagon because the company refused to allow its technology to be used in mass surveillance and fully autonomous weapons. In that situation, it was the government that wanted fewer safeguards, and the private company that wanted more.

State Laws

Whereas the Trump administration has generally held back on regulation, some states have created laws with limited ambitions. Many of the most prominent state AI laws have focused on transparency, documentation, and incident reporting. That is, the laws have been about compelling AI companies to document what they do, not to change their technology development or deployment. This approach is not accidental: transparency and incident-reporting requirements are easier to administer, easier to defend legally, and less likely to be attacked as direct interference with innovation.

One of the few laws that actually tried to change behavior was Colorado’s 2024 AI Act. It imposed requirements on AI companies to actively prevent algorithmic discrimination in consequential decisions. The law never took effect before being repealed and replaced with a narrower law focused on transparency in automated decision-making.

Further complicating state efforts have been attempts by the Trump administration to preempt state AI laws. The administration argues that a patchwork of different state laws is an incoherent and counterproductive regulatory burden. Although the administration repeatedly asked Congress to preempt state laws, those asks were rebuffed. It is currently continuing its efforts with Congress. And as a second line of attack, the administration issued a December 2025 executive order that created a Justice Department task force to sue states; the order also threatened to withhold broadband funds from states that continue to enforce their AI laws.

AI Data Centers

The local level is where public opinion has most visibly translated into action, albeit unevenly. The issue is whether local governments should allow data centers to be built nearby:

In the above Gallup poll, people against AI data centers mostly cited environmental concerns.

Normally, one should be wary of a national poll question on a “my community” local issue. By population, most of the U.S. lives in a relatively small number of cities, whereas the more likely places for data centers are the far larger number of sparsely populated rural towns. So even if urban America doesn’t want a data center, many parts of rural America do, right?

Yes and no. A data center means jobs and economic activity, so many rural areas (and, for that matter, some non-rural areas) will be enthusiastic. But not always. In January 2026, Lordstown, Ohio, population 3,300, adopted a 180-day emergency moratorium on permits for future data centers. The village council vote was unanimous after a public hearing where most speakers opposed data centers. The action was in response to a plan to build a data center nearby.

In the past year, many other towns and cities have made news by pushing back against data-center projects. But blocking a data center in one town doesn’t slow AI; it just moves the project. There appear to be enough places that want data centers—for the jobs and the tax base—to route around the places that don’t.

Americans Also Want AI

There is another kind of public opinion about AI, and it explains why so many data centers are being built. Even as Americans tell pollsters they want AI reined in, they are adopting it at a pace with few precedents. OpenAI’s ChatGPT has been the fastest-growing consumer product ever. Half of Americans use some AI tool in a typical week. People are voting with their time and their wallets, and that vote is toward more AI.

This is less a contradiction than an honest human position: people want the benefits of AI yet also want to be protected from its costs. The two desires coexist, often in the same person. Some of the worry about AI is even a backhanded compliment to how well the technology works: the same capability that makes AI worth using is what can make it feel threatening.

So while considering public opinion polls about AI, it’s important to know what’s often left unsaid: that many people concerned about AI are not opposed enough to take actions that would undermine their own benefits from AI.

What Happens Next

Given that AI’s rise has so far not been significantly affected by government policies, can we expect that to change in the foreseeable future?

Because AI is taking society into uncharted territory, the “foreseeable future” isn’t very foreseeable:

• On one hand, the factors that have enabled the relatively unregulated rise of AI remain: concern about China, fear of stifling innovation, the ability to route around localized pockets of resistance. In addition, the rapid economic expansion of the AI industry has allowed the companies involved to become major political donors and lobbying forces. The companies’ incentives are to use their newfound political influence to promote a combination of self-regulation and government regulation that serves their interests. This may also involve supporting laws or policies the companies see as inevitable, such as requiring offsets for environmental or electricity-price effects.

• On the other hand, it is conceivable that the widespread concern about AI could be the basis of a political movement broader than pockets of resistance to specific data-center projects. Especially if AI-driven layoffs accelerate visibly, diffuse worry could harden into an organized backlash. It could be fueled by fear of job loss, scarcity of entry-level jobs, rising electricity bills, environmental concerns, and resentment of concentrated wealth. Figures as far apart as Bernie Sanders and Steve Bannon are already gesturing at it. The challenge for this movement would be to avoid derailing economic growth by clamping down on AI too hard. In this spirit, Sanders recently floated a plan that emphasized participation over regulation: the government would take a 50% stake in leading AI companies so the wealth they create could be distributed more broadly. Sanders claimed it would also allow greater influence on safety, environmental, and other concerns.

Although much is unclear, the concerns Americans share about AI will keep colliding with AI’s appeal—not just its usefulness to individuals but also its potential to drive economic growth, accelerate scientific discovery, and deliver other broad benefits. How that collision between costs and benefits resolves in policy will be among the defining questions of the years ahead.