AlgoSoc / Opinion
September 17, 2026

AI insiders warn it could end humanity. Does the public agree?

Artificial intelligence is becoming an increasingly visible part of everyday life, public debate and political decision-making. Yet public views of these technologies remain far from uniform: alongside expectations of progress and convenience are concerns about control, inequality, misinformation and more fundamental societal risks.

Last week, researchers at Anthropic and OpenAI warned publicly that AI could wipe out humanity, with one putting the risk at more than 10% within a decade. The CEOs of both companies called for a slowdown, while critics suggest these doomsday warnings mainly generate hype, attract investment and push for regulation that suits the companies.

But does the public share this fear?

Public opinion matters in this debate: how citizens see the risks of AI shapes the political room governments have to regulate these technologies that could supposedly end the human race. Lucky for all of us, the AlgoSoc panel has been following the same group of Dutch adults for three years, asking about precisely these perceptions.

A public split down the middle

First, we paint a broader picture of people’s general attitudes towards AI. We asked a representative sample of the Dutch population six questions on feelings about, expectations of and preferences for AI, combined into a single attitude score based on previous academic work.

Figure 1. Overall attitude toward AI (composite of six items), spring 2026.

We see a somewhat surprising split: 46% lean positive, 41% lean negative and 12% sit in the middle. This split was stable compared with our previous round. Despite the intense hype around AI, the public has neither been completely swept up in enthusiasm nor turned against it as a whole.

Is AI a threat to humanity?

People with a negative view of AI got a follow-up question: to what extent do you believe AI is a threat to the survival of humanity?

Figure 2. “AI is a threat to the survival of humanity”, among sceptics who were asked and among all respondents.

Among these sceptics, 40% agree that AI threatens humanity’s survival. Another 28% are unsure and 33% disagree.

Across the whole population, that comes to at least one in five Dutch adults (19%). That is quite a striking number, and means that even before last week’s warnings, a substantial share of the public already took seriously a scenario that until recently seemed like science fiction.

Who is worried?

The concern about AI leading to the extinction of humanity is not spread evenly.

Figure 3. “AI is a threat to the survival of humanity” by AI chatbot use, household income and education (all respondents; those not asked counted as disagreeing).

The biggest difference is experience with AI. Among people who have never used a chatbot like ChatGPT, 30% see AI as a threat to humanity, while among regular users, the number is a third of that: only 10%. This is largely because sceptics tend not to use AI: 63% of non-users have a negative view of AI, compared with 21% of regular users. Among sceptics themselves, around four in ten see AI as a threat whether they use chatbots or not.

Income and education show a similar but weaker pattern: concerns that AI could wipe out humanity decrease with higher income and education.

What we can and can’t say yet

These findings describe public opinion before last week (May and June of 2026), not how people are reacting now. Existential warnings used to come from science fiction and a handful of researchers; now they come from inside the companies building the technology. These warnings are being communicated to a public where the worry already exists. Whether they deepen it is an open question.

We will know more soon: in October, we will ask the same people the same questions again. Because we follow the same individuals, we can test whether warnings from inside the AI industry shift public opinion, or whether it is more stable than the headlines suggest.

Want to learn more? At AlgoSoc we have created the AI Public Opinion monitor, a handy dashboard that allows anyone to explore years of public opinion data surrounding algorithms and AI. Check it out!