Summary: Anthropic’s first Public Record survey, released June 12, 2026, drew nearly 52,000 American responses from fieldwork in November and December 2025. It found curing disease led hopes at 48%, job loss led fears at 64%, 71% supported a government role, and only 15% trusted AI companies to decide AI development and use.
The baseline: Anthropic launches Public Record
The first wave of Anthropic Public Record is here. Released on June 12, 2026, the series is Anthropic’s new recurring effort to follow how the American public thinks and feels about AI. Its opening round was fielded in November and December 2025 and gathered answers from nearly 52,000 Americans — one of the biggest public-opinion studies on AI carried out so far. For anyone building AI products, the results amount to essential reading: a rare, data-rich picture of what the public expects, fears, and wants governments to do about AI.
Public Record also breaks with Anthropic’s earlier research, which either centered on Claude users (the qualitative study covering 81,000 users) or drew on anonymized usage data (the Economic Index). This survey speaks to the general public, non-users included — and that difference counts, because it records the attitudes of people who may never have opened an AI tool yet will still live with its consequences.
Where Americans want AI to deliver
Given a menu of 17 options and asked to name their three top hopes, Americans clustered around tangible, life-changing gains:
- Curing diseases such as cancer or Alzheimer’s led at 48% — a 12 percentage point cushion over the runner-up.
- Helping people with disabilities took second at 36%.
- Making technological progress and making life easier shared the next slot at 23% apiece.
Sunk at the bottom were hopes built on AI standing in for human contact, therapy or easing loneliness among them. Read one way, the public treats AI as an instrument for fixing defined problems — not as a swap-in for human relationships.
The signal for product builders is blunt: anchor your product to specific, measurable outcomes in health, accessibility, or everyday convenience, and don’t sell it as a stand-in for human interaction unless you can show solid evidence that it works.
The fear ranking: jobs, dependency, and false information
Respondents flagged concerns from a list of 20 potential harms, and the results skew toward the near-term and concrete rather than science fiction:
- Job loss led the list, with 64% of Americans worried — and it ranked first in every single state.
- Cognitive dependency (AI leaving people unable to think for themselves) came next at 56%.
- Misinformation took third at 52%.
Criminal use and surveillance surfaced often as well. Tellingly, Americans are more alarmed by AI being misused than by AI “going rogue” — in step with the observation that these fears echo earlier technologies (automation, smartphones, social media).
The job-loss data splits along two telling lines:
- Education tracks with worry: Americans holding postgraduate degrees run nearly 10 percentage points more anxious about job loss than those with a high school education or less. The most concerned workers are the ones whose jobs already overlap with what AI can do.
- Use dampens fear: Americans who work with AI every day are considerably less worried about job loss (54%) than those who never touch it (70%). First-hand experience seems to expose the technology’s limits and teach people to fold it into their work rather than be pushed aside by it.
Cognitive dependency, meanwhile, is mostly an anticipatory fear. Among the 56% who worry about it, only about 1/5 would feel significant disruption if AI disappeared tomorrow. Yet among the 44% who don’t worry, roughly 1/3 would. Actual reliance therefore sits below the level of fear — with a twist: plenty of people are already depending on AI without noticing it.
Governing AI: cross-party backing for a government role
Support for a government role in AI cuts across the political spectrum:
- 71% overall say government should be involved in AI development and regulation.
- The backing crosses party lines: 79% of Democrats, 68% of Republicans, and 69% of Independents.
- Majorities in every state and territory agree, spanning 81% in D.C. down to 63% in Hawaii.
Narrowed to specific domains, just two won majority support for more than a minimal government role: privacy (56%) and child safety (52%). National security showed the slimmest partisan gap, at just 3 points.
And what should the industry do so that AI benefits humanity? Respondents settled on two high-leverage actions:
- Hold AI companies legally liable for harm (47% placed this among their top three).
- Prioritize safety over growth (44%).
Independent watchdogs with real power (29%) and slowing AI development for safety (27%) came next.
For product teams, the message is to move first on accountability and safety: put transparent safety measures in place, publish clear liability policies, and bring in third-party audits instead of waiting for regulation to demand them. The public is paying attention.
The trust gap: AI companies at just 15%
The starkest figure may be this one: only 15% of Americans trust AI companies to decide how AI gets developed and used. That trails the federal government (20%), state and local government (19%), and international bodies (20%) — and sits far beneath independent experts at 43%.
Even the heaviest users — the roughly 6% of Americans who rely on AI daily for both work and personal life — trust AI companies only slightly more, and they back government involvement at nearly the same rate as the country overall (74% vs. 71%). Put differently, even those who gain the most from AI still want someone else keeping watch.
That gap doubles as a strategic opening. Independent experts hold 43% trust — close to three times the figure for AI companies — so partnering with academics, NGOs, or certification bodies could narrow the distance. And being open about how your models work, what data they use, and how you handle failures is no longer optional; it now reads as a competitive advantage.
How product builders should respond
Taken together, the survey describes an American public that is hopeful, fearful, and skeptical. The practical takeaways:
- Anchor to concrete value: Point your product at the hopes that rank highest — health, accessibility, an easier daily life — and skip vague “AI for AI’s sake” positioning.
- Face job-loss anxiety head-on: Even if your product displaces no one, the public fears it could. Center the story on how the tool augments people and helps them do their jobs better, not on automation that removes them.
- Own accountability early: People want legal liability for harm and safety placed ahead of growth. Bake safety in from day one, document how you work, be honest about limitations, and weigh third-party audits or independent oversight.
- Treat trust as a long game: At 15%, trust in AI companies leaves a long road ahead. Work with independent experts, publish safety research, and engage regulators before you must. Trust gets built through steady, transparent action.
The series will run again on a regular cycle and widen beyond the US, so these numbers will move. For now they set a baseline: Americans want AI to solve real problems, fear the disruption it may bring, and expect the companies behind it to answer for its effects. Products that honor those expectations — tangible benefits delivered, fears addressed, oversight welcomed — will be best placed to come out ahead.