A new POLITICO Poll study shows how quickly Americans’ attitudes toward artificial intelligence are changing. 63% of those surveyed believe that advanced AI poses at least a moderate risk of wiping out humanity. At the same time, a significant share of respondents support slowing down or temporarily pausing the development of more powerful AI models.

The poll was conducted by the independent British research firm Public First from September 13 to 15, 2026. The study was part of a regular POLITICO and Public First project examining public opinion in the United States.

Americans are ready to slow AI development

Concerns about AI’s potentially catastrophic consequences are accompanied by a more practical question: whether the development of increasingly powerful models should continue at its current pace.

The poll results show that Americans generally lean toward a more cautious approach. A significant share of respondents would prefer to slow down or temporarily pause the development of advanced AI systems, believing that existing technologies are already sufficiently advanced to deliver practical benefits.

At the same time, the public does not show a unified attitude toward further AI development. Some Americans believe that the potential economic and technological advantages justify continuing the race despite the existing risks.

The divide is especially noticeable among Donald Trump’s supporters: 44% of them would prefer to pause the further development of more powerful models, while 40% believe development should continue for the sake of possible benefits.

How much Americans trust warnings from AI companies

The poll was conducted against the backdrop of an unusual debate within the industry itself.

Executives at the largest AI companies have begun publicly speaking about the need for more cautious development of the technology. In particular, Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman have called for slowing the development of the most advanced systems.

About half of Americans believe that AI company executives warn about the dangers of artificial intelligence because they genuinely believe in the existing risks. Only 26% of respondents think such statements are a marketing tactic meant to emphasize the power and importance of the companies’ products.

At the same time, the mere existence of such warnings does not mean that experts agree on the likelihood of a catastrophic scenario. This is about how the American public perceives possible risks, not a scientific assessment of the probability of humanity’s extinction.

AI worries people for reasons beyond a hypothetical end of the world

Fear of humanity’s destruction turned out to be part of a broader wariness among Americans toward artificial intelligence.

The poll tested people’s attitudes toward a wide range of AI use cases in everyday life. In many of them, a majority of respondents said they would feel uncomfortable if artificial intelligence took on a more prominent role.

This is important for understanding the poll’s results. Public anxiety is linked not only to the hypothetical emergence of a superintelligent system. People are also concerned about the very concrete consequences of the spread of AI — its impact on work, society, and everyday life.

Similar concerns were also recorded in previous Public First studies. In one of them, respondents considered potential risks from AI — from fraud and disinformation to more extreme scenarios related to humanity’s existence.

Between safety and technological competition

At the same time, another question remains alongside the discussion of risks: how far the United States can afford to slow AI development if other countries continue moving forward.

This is precisely where one of the main conflicts around AI policy in the United States arises. On the one hand, the public shows concern about the possible consequences of overly rapid technological development. On the other hand, government and technology institutions are taking competition with China into account and are seeking to preserve American leadership in artificial intelligence.

As a result, the discussion is gradually shifting from the question “do we need AI” to a much more complex problem: what pace of development is acceptable, and what restrictions should accompany the creation of more powerful systems.

The POLITICO poll shows that the American public debate has already moved far beyond interest in new chatbot features. Artificial intelligence is increasingly seen as a technology capable of bringing significant benefits while also creating serious risks. And the question of how quickly such systems should be developed is becoming a matter of public and political debate.