Results from the first Anthropic Public Record

What happens when a leading AI company actually listens to the public? Anthropic just released results from its first “Public Record” survey — a massive online poll of nearly 52,000 Americans conducted in late 2025 — to map out genuine public attitudes toward AI. This is more than corporate PR; it is a mirror reflecting the vast trust gap between the AI industry and everyday people.

Hopes: Curing Disease, Not Replacing Humans The survey’s most striking finding is that nearly half (48%) of Americans ranked curing diseases like cancer or Alzheimer’s among their top three hopes for AI — 12 points ahead of the second-place item, helping people with disabilities (36%). Notably, the lowest-ranked hopes were those where AI might substitute for human contact, such as therapy or reducing loneliness. In other words, people welcome AI as a super-tool to tackle medical grand challenges, but they are wary of AI as an “artificial companion.” This reveals a deep psychological truth: we still want technology to serve humans, not replace them.

Fears: Beyond Job Loss, the Terror of Forgetting How to Think If hopes point toward concrete goals, fears are far messier. Job loss from AI was the top worry across every state, held by 64%. But the second-biggest anxiety may surprise many: 56% feared “cognitive dependency” — that humans might become so reliant on AI they lose the ability to think and solve problems independently. This edged out misinformation (52%), signaling that public concern has evolved from “Will AI lie to me?” to “Will AI make me dumber?” This fear strikes at the core contradiction of technological progress: we invent tools to extend our abilities, but over-outsourcing our thinking may ultimately diminish us.

Consensus: Rare Unity Across Party and Zip Code In a deeply polarized country, AI turns out to be a unifying issue. The survey found that regardless of political party, geography, or education level, Americans’ hopes and fears about AI were remarkably similar. Over 70% believe the government should play a role in regulating AI, and this support was bipartisan. The areas where people most want government action: privacy (56%), child safety (52%), and liability for harms (49%). Moreover, 47% said that holding AI companies legally liable for harm is the best way to ensure AI benefits humanity — ranking even higher than “prioritizing safety over growth” (44%).

Trust Crisis: Only 15% Believe in AI Companies Perhaps the most glaring number is this: only 15% of Americans trust AI companies to make decisions about how AI is developed and used. That means 85 out of every 100 average citizens do not trust firms like OpenAI or Anthropic to self-govern. This explains the overwhelming public demand for external regulation and accountability. When companies tout self-regulation, the people have already voted with their distrust.

What This Foretells First, the “source of legitimacy” for AI governance is shifting from industry self-policing toward public mandate. Companies that continue to see regulation as an obstacle will only deepen the trust gap. Second, “cognitive dependency” as a newly prominent fear could spur a wave of design principles centered on human agency — for instance, AI systems that default to preserving human decision points rather than automating everything. Third, the survey itself is a signal: top AI firms are now systematically measuring public sentiment, which will become foundational data for future policy battles.

For builders, this report is not a threat assessment but a user requirements document. People are willing to pay for an AI that cures cancer, but they will reject an AI that takes away jobs and atrophies their thinking. If we want to build AI that is broadly accepted, we cannot just bury our heads in training bigger models — we must also lift our heads and listen to the silent majority.