Anthropic went past its own user base and asked the general public a simple question: what do you actually think about AI? The answers are now out, and they’re worth your attention. According to Anthropic, the first wave of its new survey series, the Anthropic Public Record, reached nearly 52,000 Americans in November and December of 2025. This is the first time the company has gone looking for non-users, people who don’t touch Claude or ChatGPT but still have strong opinions about where this technology is headed.
What stands out here is the gap between hope and fear. People want big things from AI, and they’re scared of it at the same time.
What people hope for
The optimism clusters around health and human benefit, not productivity or profit. The top hopes Americans ranked in their top three:
- Curing diseases like cancer or Alzheimer’s: 48%
- Helping people with disabilities: 36%
- Making technological progress: 23%
- Making life easier in general: 23%
Nearly half the country sees AI primarily as a path to medical breakthroughs. That’s a different story than the one the industry usually tells about coding agents and office automation.
What people fear
The fears are more unified, and more pointed. Job loss topped the list in every single state Anthropic surveyed. The breakdown:
- AI-induced job loss: 64%
- Cognitive dependency (relying on AI to think): 56%
- Misinformation: 52%
That second one is interesting. More than half of Americans worry not just about losing work, but about losing the habit of thinking for themselves. That’s a fear the AI industry has barely addressed, and it’s sitting at 56%.
The regulation signal
Here’s the number that should make every AI company pay attention: over 70% of Americans surveyed believe the government should play a role in regulating AI. And this support was bipartisan. In a country that agrees on almost nothing, AI oversight is a rare point of consensus across party lines.
This matters because it reframes the regulation debate. It’s often framed as Washington versus Silicon Valley, or one party versus another. The data says it’s neither. It’s a broad public expectation cutting across the usual divides.
Why the methodology matters
Most AI sentiment research surveys people who already use the tools. That skews everything toward enthusiasts. Anthropic’s approach here deliberately reached non-users and measured how attitudes shift across demographic lines. That’s what makes the job-loss and cognitive-dependency numbers credible: they include the voices of people who are watching AI from the outside and don’t like what they see.
One limitation to keep in mind: this is a single snapshot from late 2025, and self-reported attitudes about a fast-moving technology can change quickly. Anthropic frames this as the first wave, which suggests more data is coming to track how these views move over time.
What you can do with this
If you build, sell, or write about AI, this is a map of the conversation your audience is already having:
- Lead with health and human benefit, not raw efficiency. That’s where public hope actually lives.
- Take the cognitive-dependency fear seriously. Tools that make people feel sharper, not more dependent, are addressing a real and widespread worry.
- Stop treating regulation as a fringe position. A 70% bipartisan majority wants it. Products and messaging that ignore that are out of step with the public.
- Remember the non-users. The loudest fears come from people who aren’t in your funnel yet, and they vote, hire, and shape policy.
The headline finding is the tension itself. Americans want AI to cure cancer and fear it will take their jobs and dull their minds, all at once. That contradiction is the real public mood, and Anthropic just put hard numbers on it. The next wave will tell us which way the needle moves.
Full results are available in the Anthropic Public Record results.