America's AI mood: hopeful, anxious, and short on trust
Anthropic ran a nationally representative survey of nearly 52,000 Americans (fielded in late 2025) to map how the public - not just Claude users - feels about AI. The headline: people are eager for the upside but nervous about the disruption, and they want the companies building AI held accountable.
What people hope for
Asked to pick their top three hopes from a list of 17, respondents led with curing diseases like cancer and Alzheimer's (48%), followed by helping people with disabilities (36%), then general technological progress and convenience. A telling detail: hopes that AI might substitute for human contact - therapy, easing loneliness - ranked lowest.
What people fear
The fears skew near-term and concrete rather than science-fiction:
- Job loss was the runaway top concern at 64%, and it topped the list in every single state.
- Close behind sat cognitive dependency - losing the ability to think for yourself - at 56%.
- Then came misinformation, at 52%.
Interestingly, Americans worried more about AI misuse (criminal use, surveillance, terrorism) than about AI "going rogue."
The usage paradox
One of the most striking patterns: hands-on familiarity cools the fear. People who use AI daily at work were notably less worried about job loss (54% vs. 70% for non-users) and dependency (46% vs. 62%). Whether that's because using the tools reveals their limits, or because early adopters are simply sunnier, is an open question - but the gap is large and consistent.
Another counterintuitive twist: cognitive dependency looks more anticipatory than real. Of those worried about it, only about a fifth said they'd feel significant disruption if AI vanished tomorrow - actually lower than among people who claim not to worry about it.
The trust deficit, and what voters want done about it
Here's the number that should sober every lab, Anthropic included: just 15% of Americans trust AI companies to make decisions about how the technology is built and used - the lowest of any institution tested, well behind independent experts (43%).
- Over 70% want government involved in regulating AI, a bipartisan supermajority that holds in every state.
- They most want action on privacy, child safety, and liability for harm.
- Asked how to keep AI beneficial, the top picks were holding companies legally liable (47%) and prioritizing safety over growth (44%).
Why Anthropic is publishing this
The survey is part of a broader push - alongside Anthropic Interviewer and the Economic Index - to fold public input into AI development, and it dovetails with the policy frameworks Anthropic released just days earlier. The subtext is pointed: if the public barely trusts AI firms to govern themselves, the case for outside guardrails rather writes itself. Anthropic plans to repeat the survey and eventually expand beyond the US.