The artificial intelligence industry has spent the last several years telling people what they need. A massive new survey suggests the public has been trying to tell the industry what it wants — and the two visions don’t align nearly as well as Silicon Valley would like to believe.
The largest-ever public survey on AI attitudes, conducted by researchers at the University of Queensland and published in collaboration with the Varieties of Democracy Institute, collected responses from more than 17,000 people across 21 countries. The findings, reported by The Next Web, paint a picture of a global public that is far more cautious, far more specific, and far more skeptical about AI than the companies building it seem to appreciate.
The headline finding is stark. People overwhelmingly want AI that serves practical, bounded purposes — healthcare diagnostics, scientific research, disaster response — and they’re deeply uncomfortable with AI systems that make autonomous decisions about their lives. Not somewhat uncomfortable. Deeply.
This isn’t a survey of Luddites. Respondents across nearly every country and demographic expressed genuine enthusiasm about AI’s potential in specific domains. Medical applications ranked consistently high. So did environmental monitoring and scientific discovery. These are areas where AI acts as a tool wielded by human experts, augmenting their capabilities rather than replacing their judgment. The pattern is unmistakable: people want AI as an instrument, not an agent.
What they don’t want is equally revealing. Autonomous weapons systems provoked near-universal opposition. AI-driven surveillance drew sharp resistance across democratic and non-democratic countries alike, though the intensity varied. And the deployment of AI in criminal justice — sentencing recommendations, predictive policing, parole decisions — generated profound anxiety even among respondents who were otherwise optimistic about the technology.
The researchers, led by Dr. Chinmayi Arun and her team at the University of Queensland, designed the survey to capture not just binary approval or disapproval but the texture of public opinion: what conditions people attach to their acceptance, what trade-offs they’re willing to make, and where their red lines fall. The methodology spanned countries as varied as the United States, India, Germany, Nigeria, Brazil, and Japan.
One of the most striking dimensions of the data involves trust. Or rather, the absence of it.
Respondents expressed low confidence in technology companies to self-regulate. They expressed low confidence in governments to regulate effectively. And they expressed low confidence that their own preferences would matter in shaping how AI gets deployed. This triple deficit of trust didn’t vary as much by country as one might expect. It was remarkably consistent. Citizens of wealthy democracies and citizens of developing nations shared a common intuition: the decisions about AI’s future are being made without them, and probably against their interests.
The Gap Between What’s Being Built and What’s Being Asked For
This finding should unsettle executives at every major AI company. The industry’s current trajectory — toward increasingly autonomous systems, toward AI agents that book your flights and manage your calendar and negotiate on your behalf — represents a bet that people will come around once they experience the convenience. The survey suggests the opposite may be true. People aren’t ignorant of what AI can do. They’ve formed opinions about what it should do. And those opinions are more conservative than the product roadmaps coming out of San Francisco and Shenzhen.
Consider the specific case of generative AI. The survey was conducted during a period when ChatGPT, Midjourney, and similar tools had already achieved mass awareness. Respondents didn’t reject these tools outright. But they drew sharp distinctions between using AI to assist with creative work and using AI to replace creative workers. They wanted the tool. They didn’t want the displacement.
This tracks with recent labor market data and ongoing disputes in industries from Hollywood to journalism. The Writers Guild of America strike in 2023 centered partly on AI guardrails. Visual artists have filed lawsuits over training data. Musicians have pushed back against AI-generated vocals mimicking their voices. The survey suggests these aren’t fringe concerns but mainstream ones, shared by people who have no personal stake in creative industries.
There’s a geographic dimension worth examining. Respondents in countries with stronger data protection regimes — Germany, for instance — tended to express more specific and granular preferences about AI governance. They didn’t just say they wanted regulation; they specified what kind. Transparency requirements. Algorithmic auditing. Mandatory human oversight for high-stakes decisions. Meanwhile, respondents in countries with weaker institutional frameworks expressed the same underlying concerns but with less confidence that any regulatory mechanism could actually work.
The United States occupied an interesting middle position. American respondents were among the most enthusiastic about AI’s economic potential and simultaneously among the most distrustful of both corporate and governmental oversight. This paradox — wanting the benefits, trusting nobody to manage the risks — may help explain the chaotic state of American AI policy, which has oscillated between executive orders promoting AI development and executive orders attempting to constrain it, depending on who occupies the White House.
India’s respondents stood out for their relative optimism about AI in public services, reflecting perhaps both the country’s rapid digitization and its experience with large-scale technology-driven governance programs like Aadhaar. But even Indian respondents drew lines at autonomous decision-making in areas affecting individual rights.
The survey also probed something that most AI discourse ignores: what people think about the concentration of AI power in a small number of companies. The results were unambiguous. Across every region, respondents expressed concern about the degree to which a handful of firms — primarily American and Chinese — control the development and deployment of AI systems that affect billions of people. They didn’t frame it in antitrust language. They framed it in terms of accountability. Who do you complain to when an algorithm denies your loan application? Who is responsible when an AI medical diagnostic is wrong?
These aren’t hypothetical questions anymore. They’re happening now, in real systems, affecting real people. And the survey reveals that the public knows it.
What makes this research particularly valuable is its scale and rigor. Previous surveys on AI attitudes — conducted by Pew, Ipsos, the Ada Lovelace Institute, and others — have typically covered fewer countries, smaller samples, or narrower question sets. The University of Queensland study’s 17,000-respondent, 21-country design gives it a statistical heft that’s harder to dismiss. As The Next Web noted, this is the largest survey of its kind ever conducted.
And yet the industry’s response to public opinion research has historically ranged from polite acknowledgment to outright indifference. OpenAI, Google DeepMind, Anthropic, and Meta have all published responsible AI principles and funded governance research. But the gap between stated principles and shipping decisions remains wide. Products launch fast. Governance frameworks arrive later, if they arrive at all.
The survey’s implications for policymakers are equally pointed. Respondents didn’t just want regulation — they wanted specific kinds of regulation, and they wanted it to be enforceable. The European Union’s AI Act, which began taking effect in stages in 2024, represents the most comprehensive attempt to codify these preferences into law. But even the EU’s approach has drawn criticism for being simultaneously too broad in some areas and too narrow in others, a reflection of how difficult it is to translate public sentiment into workable statute.
There’s a deeper tension the survey exposes, one that the AI industry has mostly avoided confronting directly. The public doesn’t share the industry’s foundational assumption that more capable AI is inherently better. People want AI that is more useful, more reliable, and more accountable. Capability without accountability reads as threat, not progress. And the survey data suggests this isn’t a communication problem that better marketing can solve. It’s a values gap.
The researchers recommend that AI developers incorporate public input mechanisms into their design and deployment processes — not as a public relations exercise but as a substantive constraint on what gets built and how. They also recommend that governments invest in AI literacy programs, not to make people more accepting of AI but to make them more effective participants in governance debates.
Whether any of this will change the industry’s direction is an open question. The incentive structures in AI development — venture capital timelines, competitive pressure from China, the race to artificial general intelligence — all push toward speed and capability, not deliberation and restraint. But the survey offers a data point that’s hard to argue with: 17,000 people, 21 countries, and a message that is consistent across nearly all of them.
Build AI that helps us. Don’t build AI that decides for us. And for the love of everything, let us have a say.
The industry would do well to listen. The question is whether it’s structurally capable of doing so.
WebProNews is an iEntry Publication