Most Americans see serious risks from advanced artificial intelligence, poll finds

Yesterday 21:45

By: Azzat Manal

Most Americans see serious risks from advanced artificial intelligence, poll finds

A new poll indicates that a majority of Americans believe advanced artificial intelligence could pose a serious long-term risk to humanity, underscoring growing public concern as the technology develops at a rapid pace.

The survey, conducted online by London-based Public First for Politico from September 13 to 15 among 2,064 U.S. adults, found that 63% of respondents considered there to be at least a moderate risk that advanced AI could eventually destroy humanity. The poll has a margin of error of plus or minus 2.2 percentage points.

Within that group, 17% described the possibility as almost certain, while 20% considered the risk significant and another 26% characterized it as moderate. The findings point to a substantial level of concern among Americans about the potential consequences of increasingly advanced AI systems, although they reflect perceptions rather than a scientific estimate of the actual probability of human extinction.

The survey also found that many respondents favor a slower approach to the development of increasingly capable AI. Some 48% said current AI capabilities were sufficient and that development should be paused to allow risks to be reduced, while 31% preferred continued development because of the potential economic and technological benefits of more advanced systems.

Public concern comes amid an increasingly active debate within the technology sector about the pace at which frontier AI models should be developed. Several prominent industry figures have recently warned that increasingly autonomous and capable systems could create risks that are difficult to manage if safety measures do not keep pace with technological progress.

Among the issues discussed by AI researchers and executives are the possibility of losing meaningful human control over highly capable systems, malicious use of AI for cyberattacks or biological threats, and the broader economic and social consequences of automation. The likelihood and timing of extreme scenarios remain subjects of significant disagreement among researchers, with no established scientific consensus that advanced AI will destroy humanity.

The poll also examined how Americans interpret warnings from technology leaders. Around half of respondents said they believed AI executives who warn about catastrophic consequences genuinely hold those concerns. At the same time, 26% suspected that such warnings could partly serve a marketing purpose by making AI products appear more powerful or consequential.

The results come as the United States continues to debate how to balance technological innovation with safety and oversight. Policymakers and industry leaders face competing pressures: maintaining investment and international competitiveness while addressing concerns about privacy, cybersecurity, misinformation, labor disruption and the potential risks associated with future AI systems.

Recent statements from major technology figures have added momentum to the discussion. Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman and Elon Musk have all been associated with calls for greater caution over the development of advanced AI, although their positions and proposed approaches differ.

The findings also illustrate the gap that can emerge between technological development and public confidence. AI systems are increasingly being incorporated into workplaces, consumer services, research and other areas of daily life, while questions about their long-term implications remain unsettled.

As the industry moves toward more capable models, the debate is likely to focus increasingly on the safeguards needed to manage potential risks without preventing beneficial applications. For the public, the latest polling suggests that concerns about the most extreme consequences of AI have become a significant part of that conversation.