Public Concern Over AI Existential Risk: What the Polls Show

AI and the Shadow of Existential Risk

For decades, artificial intelligence has been framed as a tool of progress: a way to automate tasks, improve decision-making, and unlock new scientific frontiers. But in recent years, a different narrative has gained traction—one in which AI is not just a powerful technology, but a potential existential threat. What was once the domain of science fiction and niche academic debates is now entering mainstream public consciousness. Surveys from multiple independent research groups now suggest that a substantial minority of people believe AI could lead to human extinction, and that a majority support strong regulatory measures to prevent that outcome.

How Much Do People Actually Believe AI Could End Humanity?

One of the most striking findings from recent public opinion research is that many people assign a non-trivial probability to the idea that advanced AI could cause human extinction. In one nationally representative U.S. survey, respondents were asked to estimate the likelihood that AI would cause human extinction by the year 2100. The results show a highly skewed distribution: while the most common individual estimate clusters around 1%, the average (mean) probability assigned across the population is around 26%. The median estimate is about 15%, meaning half of adults think the risk is 15% or higher.

This is not a fringe belief held by a tiny handful of people. When asked about extinction risk over the next 50 years, roughly 22% of U.S. adults say they consider it at least moderately likely. Over the next decade, that figure is lower but still notable, at about 9%. These numbers suggest that while most people do not expect AI to cause extinction in the near term, a significant portion of the population sees it as a serious long-term possibility.

Worry vs. Belief: Separating Fear from Probability

Believing something is possible is not the same as worrying about it every day. Research shows that while many people think AI extinction is at least somewhat plausible, it does not yet dominate their daily concerns. When asked how much they worry about the negative effects of AI in their personal lives and in society, most people report only moderate or low levels of anxiety. This suggests that the risk is perceived more as a distant, abstract possibility than an immediate, pressing danger.

However, there is a clear relationship between belief and concern. People who think AI is more likely to cause extinction or achieve superhuman intelligence tend to report higher levels of worry. In particular, the belief that AI could cause human extinction within the next 50 years is more strongly associated with worry than the belief that AI will simply become very intelligent. This indicates that it is the catastrophic potential of AI, not just its general advancement, that drives public unease.

Where Does AI Rank Among Existential Threats?

When people are asked to name the biggest existential threats to humanity, AI does not usually top the list. Climate change, nuclear war, and pandemics are more commonly cited as the most serious dangers. In one survey, only a small minority selected AI-caused extinction as the most substantial existential threat facing humanity.

Yet AI still features prominently in the broader landscape of catastrophic risks. A significant minority of people—around 30% in some polls—see advanced AI trying to take over or destroy human civilization as an important risk. About 28% believe it is a real possibility that advanced AI could cause a breakdown in human civilization within the next fifty years. These figures suggest that while AI is not the dominant concern, it is now firmly part of the public’s mental map of global catastrophic risks.

Support for Regulation and Slowing Down AI Development

Perhaps the most policy-relevant finding is that public concern about AI risks translates into strong support for regulatory action. In a major U.S. poll, 86% of voters said they believe AI could accidentally cause a catastrophic event, and 70% agreed that mitigating the risk of extinction from AI should be a global priority, on par with risks like pandemics and nuclear war.

Support for concrete measures is even more striking:

• 72% of voters prefer slowing down the development of AI, compared to just 8% who want to speed it up.
• A large majority support federal regulation of AI, including limits on the most powerful systems.
• There is broad cross-partisan agreement on these points, with majorities of Democrats, Republicans, and independents all favoring caution and oversight.

This level of consensus is unusual in today’s polarized political environment and suggests that AI risk is not being treated as a partisan issue in the public mind. Instead, it is being framed as a shared challenge that requires careful management, similar to how society approaches nuclear technology or biotechnology.

Why Is Public Concern Higher Than Experts Expected?

For years, many experts and commentators assumed that the general public would be largely unaware of or indifferent to AI extinction risk. The idea that AI could cause human extinction was often dismissed as an inside-the-bubble concern of a small group of researchers and effective altruists. Recent polling shows that this assumption was wrong.

Several factors may explain why public concern is higher than expected:

• Media coverage: High-profile warnings from AI researchers, tech leaders, and public intellectuals have brought existential risk into mainstream discourse.
• Visible AI capabilities: The rapid progress of large language models and other systems has made the idea of superintelligent AI feel more concrete and less abstract.
• Broader anxiety about technology: Many people already worry about automation, job loss, misinformation, and surveillance; AI extinction risk can feel like an extreme version of these existing concerns.
• “Wisdom of the crowds” effect: When people are asked to estimate how many others believe in AI extinction risk, their guesses are often higher than their own stated beliefs, suggesting that they perceive the risk as more widely accepted than they admit in their own responses.

How Public Perception Compares to Expert Estimates

It is instructive to compare public estimates of AI extinction risk with those of experts. Surveys of AI researchers suggest that the median expert assigns a roughly 5–10% chance to AI causing human extinction. Some prominent figures give much higher estimates (e.g., above 50%), while others are more optimistic (e.g., below 5%).

The public’s average estimate of around 26% is higher than the median expert estimate, but this is largely driven by a long tail of people who assign very high probabilities. The median public estimate of about 15% is closer to the upper end of expert estimates. This suggests that while the public as a whole is somewhat more pessimistic on average, the central tendency of public belief is not wildly out of line with expert opinion.

Implications for Policy and Advocacy

The fact that a large majority of the public supports slowing down AI development and regulating powerful systems has important implications for policymakers. It means that:

• There is strong democratic legitimacy for enacting strict AI safety regulations.
• Politicians who ignore or dismiss AI extinction risk may be out of step with voter sentiment.
• Advocacy efforts focused on catastrophic and existential risks are likely to resonate with the public, not alienate them.

At the same time, the relatively low level of day-to-day worry about AI suggests that public attention is still fragile. Sustained engagement—through clear communication, transparent risk assessment, and visible regulatory action—will be needed to maintain support for long-term safety measures.

Challenges in Communicating AI Risk

One of the biggest challenges in discussing AI extinction risk is avoiding both alarmism and complacency. On one hand, overstating the immediacy of the threat can lead to panic, misinformation, and counterproductive policy responses. On the other hand, downplaying the risk can leave society unprepared for potentially catastrophic developments.

Effective communication should:

• Clearly distinguish between near-term harms (e.g., job displacement, bias, misinformation) and long-term existential risks.
• Acknowledge uncertainty: we do not know exactly how likely AI extinction is, but we do know that the stakes are extremely high.
• Emphasize that risk mitigation is not about stopping AI, but about ensuring it is developed safely and responsibly.

What the Data Suggests About the Future of AI Governance

The polling data paints a picture of a public that is cautious, risk-averse, and supportive of strong oversight. This creates a favorable environment for:

• International cooperation on AI safety, similar to nuclear non-proliferation or climate agreements.
• Establishing independent regulatory bodies with real authority over the most powerful AI systems.
• Requiring transparency, safety testing, and “off switches” for frontier AI models.
• Investing in research on AI alignment, robustness, and control mechanisms.

It also suggests that the AI industry’s current trajectory—prioritizing speed and scale over safety and caution—may be increasingly at odds with public values. Companies and governments that ignore this shift in public opinion risk losing legitimacy and facing stronger regulatory backlash in the future.

Key Takeaways from the Polling

To summarize the main findings from recent public opinion research on AI extinction risk:

• A significant minority of people believe AI could cause human extinction, especially over longer time horizons (e.g., 50 years).
• Public concern is higher than many experts previously assumed, with strong support for slowing development and increasing regulation.
• While AI is not the top existential worry for most people, it is now widely recognized as a potential catastrophic risk.
• There is broad cross-partisan support for treating AI extinction risk as a global priority, alongside pandemics and nuclear war.
• Effective policy will require balancing public concern with technical realism, avoiding both overreaction and underreaction.

Frequently Asked Questions

What percentage of people think AI could cause human extinction?

Surveys suggest that around 22% of U.S. adults think AI-caused extinction is at least moderately likely within the next 50 years, and about 9% think it is at least moderately likely within the next 10 years. When asked to give a percentage chance of extinction by 2100, the average estimate is around 26%, though the median is about 15%.

Do most people support regulating AI?

Yes. Major polls show that large majorities of voters support federal regulation of AI and believe that mitigating the risk of extinction from AI should be a global priority. In one survey, 70% agreed that preventing AI-driven extinction should be treated as seriously as other major global risks.

Is public concern about AI extinction risk overblown?

Whether the level of concern is “overblown” depends on one’s assessment of the actual risk. The public’s average estimate is higher than the median expert estimate, but the median public estimate is in the same general range as some expert estimates. Given the potentially infinite stakes of human extinction, even a small probability can justify serious precautionary measures.

Why do people worry about AI extinction if it seems so far-fetched?

People worry because the consequences of being wrong are so severe. Even if the probability is low, the potential for irreversible, civilization-ending outcomes makes the risk worth taking seriously. This is similar to how society treats other low-probability, high-impact risks like nuclear war or asteroid impacts.

What should governments do in response to public concern about AI risk?

Governments should:

• Establish independent regulatory bodies with authority over frontier AI systems.
• Require safety testing, transparency, and accountability for high-risk models.
• Invest in research on AI alignment, robustness, and control.
• Engage in international cooperation to prevent a dangerous AI arms race.
• Communicate clearly with the public about risks and safeguards, avoiding both alarmism and complacency.

Written by

Sneha Tete

• Integrated M.A. in Applied Linguistics, University of Hyderabad
• 5+ years in editorial, instructional, SEO, and research-based writing

Sneha is a relationships and lifestyle writer with a strong foundation in applied linguistics and certified training in relationship coaching. She brings over five years of writing experience to theskillvault,  crafting thoughtful, research-driven content that empowers readers to build healthier relationships, boost emotional well-being, and embrace holistic living. Her writing and insights reflect the belief that beauty, happiness, and wellness…