# Only 15% of Americans trust AI companies: inside the first Anthropic Public Record

Heya, it's Shiichan! Today I've got a news story packed with numbers and a bit more social than usual. AI is handy, but what do people really think about it deep down? A survey asked exactly that, at the scale of 50,000+ people!

## What was announced?

Anthropic's News published the results of the first wave of a new survey series called the "Anthropic Public Record." Unlike earlier research that focused on Claude users, this one measures how the general American public — including people who don't use AI — sees AI.

The first wave ran in November–December 2025 through YouGov, asking 51,993 Americans. It's a nationally representative online survey, weighted to the US Census. Anthropic plans to repeat it regularly and eventually expand beyond the US.

## Why it matters

Conversations about AI tend to be dominated by the companies and researchers building it. This survey instead surfaces what ordinary people living in society hope for and fear — across party, region, and education. And with 50,000+ respondents, you can even see differences between states.

## What changes

First, people's hopes. Picking their top three from 17 options, the most common was curing diseases (like cancer and Alzheimer's) at 48%, followed by helping people with disabilities at 36%, and general technological and life improvements at 23%.

The fears are just as clear. Counting "somewhat worried" or higher, fear of losing jobs led by far at 64%, and it barely moved across states (57–71%) — a shared worry nationwide. Next came cognitive dependency ("relying on AI so much you stop thinking for yourself") at 56%, and misinformation at 52%.

What's interesting: people's fears pointed at concrete, near-term harms rather than sci-fi risks like AI going rogue.

And here's the number that hit hardest. Asked who they'd trust to make AI development decisions, only 15% trusted AI companies — the lowest of anyone tested. That's below the federal government (20%), state/local government (19%), and international bodies (20%), and far below independent experts (43%).

## Dive Deep

Let's look a bit closer.

Fear of job loss actually grows with education: people with postgraduate degrees worried about 10 points more than those with a high school education or less. By AI exposure, 70% of non-users were worried, dropping to 54% among people who use AI daily at work. Hands-on use may build a sense of augmentation rather than replacement.

Cognitive dependency is fascinating too: 56% worried about it, yet only about one-fifth said they'd be seriously disrupted if AI suddenly vanished. Meanwhile, among the 44% who weren't worried, about one-third actually would be — perception and reality are a little twisted.

On capability, respondents rated AI across 14 workplace tasks. The highest was research at 75% ("as good or better" than humans); the lowest was service and support at 44%. Even so, nearly half of people said they'd rather AI stay out of their own work.

Regulation got specific too. 71% supported government involvement (Democrats 79%, Republicans 68%, Independents 69%). The top priority areas were privacy (56%), child safety (52%), and liability for harm (49%). Asked what keeps AI serving humanity, legal liability for harm (47%) and prioritizing safety over growth (44%) came out on top.

Anthropic sums it up like this:

> Americans are eager to realize AI's promised benefits but fear the disruption it may bring, and they want accountability from the companies building it.

By the way, this survey pairs with other efforts, like a qualitative study of 81,000 Claude users via the Anthropic Interviewer, and the Anthropic Economic Index.

## Wrap-up

- Anthropic's News published the first "Anthropic Public Record," asking 51,993 general Americans their honest views on AI
- Top hope was curing diseases (48%); top fear was job loss (64%)
- Only 15% trust AI companies — far below independent experts (43%)
- Job worry rises with education and is highest among non-users (70% vs 54%)
- 71% support government involvement, with strong agreement across parties

If you love tracking public opinion through numbers, if you think about AI policy and governance, or if you've ever wondered "what does everyone really think about AI?", this one's for you!