Anthropic published the results of its first Public Record on June 12, 2026 — a nationwide survey designed to take the pulse of what Americans actually want from artificial intelligence. The results are striking: overwhelming hope that AI will tackle humanity’s hardest problems, paired with deep distrust of the companies building it.
The Numbers: Curing Disease Tops the List
When asked to rank what they most want AI to accomplish, 48% of respondents put “curing diseases like cancer and Alzheimer’s” in their top three — a full 12 percentage points ahead of the second-ranked item, “helping people with disabilities” at 36%.
The rest of the top five:
- Therapy and reducing loneliness — 31%
- Tutoring and personalized education — 29%
- Scientific discovery and climate research — 27%
These priorities reflect a public that sees AI as a tool for solving concrete, human-scale problems rather than an abstract productivity booster. Notably absent from the top tier: coding assistants, enterprise automation, or creative tools — the very use cases that dominate industry discourse.
The Trust Gap
If the optimism is striking, the trust deficit is sobering. Only 15% of respondents said they trust AI companies to make key decisions about how the technology is developed and deployed. This figure held roughly steady across demographic lines — it’s not a partisan or generational split, but a broad societal skepticism.
This trust gap creates a paradox for companies like Anthropic: the public wants AI to cure disease and transform education, but doesn’t trust the companies building it to decide how. It’s a dynamic that helps explain why Anthropic has invested so heavily in its constitutional AI approach, third-party safety testing, and now, this very Public Record exercise.
Jobs Anxiety Remains Real
Alongside the hope, fear of economic displacement runs deep. A significant share of respondents expressed concern about AI-driven job losses — particularly in knowledge-work sectors that felt insulated just two years ago. This anxiety cuts across income and education levels, suggesting it’s not just about low-skill automation anymore.
Why This Matters Now
The Public Record lands during the most turbulent week in Anthropic’s history. Just three days before publishing these results, the company launched Claude Fable 5 — its most capable publicly available model ever — only to have the U.S. government order it suspended on June 12 over export control and national security concerns.
That juxtaposition — a company simultaneously publishing a public survey about AI’s benevolent potential while being forced by the government to disable its own models — crystallizes the moment AI finds itself in. The public wants the cure. The government wants the brakes. And the companies are caught in between.
Anthropic has framed the Public Record as an ongoing exercise, not a one-off. If the company follows through, this could become a meaningful feedback loop between what the public wants and what frontier labs actually build. But with trust at 15%, there’s a long road ahead.
The full Public Record results are available at anthropic.com/news/anthropic-public-record.