Half the Public Wants To Pause AI. Would Regulation Actually Make It Safer?

As leaders from across the globe gather in New York to discuss matters of national security, public opinion is turning against AI. About half of adults across six Western countries say “AI is now good enough” and support a pause in advancement to reduce risks, according to a new Politico poll.

The poll, which was conducted by Public First, asked more than 2,000 respondents in each of the six countries—the United States, Canada, the United Kingdom, France, Spain, and Germany—what they thought about the potential risks associated with advanced AI. In the United States, 78 percent of respondents say there is a risk that advanced AI could “destroy humanity,” with only 7 percent saying there is no risk at all. The majority of Americans (51 percent) say that AI industry leaders argue that “there is a real risk that AI will kill us all” because they genuinely believe it.

In each country surveyed, more respondents favored pausing development of more advanced AI than continuing it—51 percent in the United Kingdom, 50 percent in Canada, 48 percent in France and the United States, 47 percent in Spain, and 45 percent in Germany.

Attitudes shift slightly when the prospect of falling behind other countries is presented. In the U.S., 43 percent of respondents said their country should be willing to fall behind other countries in the development of AI, while 37 percent said America should not allow itself to fall behind. A majority of respondents in Britain (53 percent) and Canada (53 percent) said their countries should be willing to fall behind internationally, while only 43 percent of respondents in France, 45 percent in Germany, and 42 percent in Spain agreed.

While there is little reason to believe an AI apocalypse is imminent, high-profile CEOs, such as Sam Altman and Dario Amodei, have joined calls for government regulation to curb what they see as a threat. Sen. Bernie Sanders (I–Vt.) and Rep. Greg Casar (D–Texas) have been prominent voices in support of AI regulation, announcing the Ban Artificial Superintelligence Act on Wednesday, which would ban highly advanced AI and establish a regulatory agency. But President Donald Trump has thus far resisted those calls, saying he refuses “to stifle growth of something that will be bigger than the industrial revolution.”

Of course, there is nothing stopping AI companies from slowing down themselves. OpenAI has already delayed parts of its Astra model as it strengthened safeguards, while Anthropic has published a Responsible Scaling Policy to manage risks, explicitly committing to restricting deployment or further scaling if a model crosses a dangerous capability threshold before adequate safeguards are ready. Similarly, Google created the Frontier Safety Framework—”a set of protocols for proactively identifying future AI capabilities that could cause severe harm and putting in place mechanisms to detect and mitigate them.”

There is no reason to believe that new government regulations on AI would eliminate risk. Indeed, if future regulations cap AI developers’ liability, provide government-backed insurance, or create an official stamp of approval for models that satisfy prescribed and arbitrary safety tests, companies could retain much of the upside from developing increasingly powerful systems while shifting some of the downside onto taxpayers or the wider public.

When Altman and other AI leaders ask for government intervention, the possibility of regulatory capture cannot be dismissed. The companies calling most loudly for AI regulation are also among those best equipped to comply with it. Licensing regimes, mandatory audits, reporting requirements, and expensive safety standards impose fixed costs that established firms such as OpenAI, Anthropic, and Google can absorb more easily than smaller competitors. Altman has urged Congress to consider licensing or registration requirements. That does not mean industry warnings about AI risk are disingenuous, but advocating for regulation does allow these executives to present themselves as responsible stewards of a dangerous technology while also supporting rules that could simultaneously raise barriers to entry and entrench their position in the market.

Interestingly, the Politico poll suggests younger Americans are more likely to use AI regularly, and yet they are among the most cautious cohorts. Of respondents aged 18–24, 57 percent support pausing the development of more advanced AI, while just 24 percent favor continuing it. Among Americans aged 65 and older, by comparison, 47 percent support a pause and 33 percent favor continued development. But of the 18–24 cohort, 55 percent use AI at least weekly, compared to just 26 percent of over-65s.

Some of this AI anxiety among the young is likely driven by fears regarding job security. According to the Pew Research Center, 73 percent of adults under 30 think that AI will lead to fewer jobs, up from 61 percent two years ago.

There is little evidence that AI is causing mass unemployment thus far. Using payroll data covering millions of U.S. workers from January 2021 to June 2026, the Stanford Digital Economy Lab found “no evidence of widespread, economy-wide job displacement.” While employment of younger workers (22- to 25-year-olds) in AI-exposed occupations is 19 percent below where it would be if it had followed the same trajectory as their less-exposed peers, “experienced workers show no comparable gap.” Researchers at Yale’s Budget Lab, analyzing data from ChatGPT’s release in November 2022 to August 2025, found that “the broader labor market has not experienced a discernible disruption since ChatGPT’s release.” They also found little change in the proportion of workers in occupations exposed to AI. Indeed, an April Census Bureau report covering November 2025 to January 2026 found that “AI-related employment decreases are rare, occurring in only 2% of firms.”

For now, there is little that can be done to eliminate public anxiety about AI. But the government should be wary of treating regulation as an easy fix for risks that are still poorly understood.