Americans fear AI job loss, but daily users worry less

Silicon Valley loves to debate whether AI will birth digital gods or bring utopia, but outside tech bubbles, real-world sentiment looks very different.

Anthropic just released findings from its inaugural Anthropic Public Record survey, polling nearly 52,000 Americans across late 2025. Conducted with YouGov, it represents one of the largest snapshots of mainstream public perception around AI to date.

Here is what Americans actually want from AI, what they dread, and who they blame.

Big hopes, bigger job fears

When asked what they hope AI accomplishes, Americans overwhelmingly pointed to healthcare. Nearly half (48%) placed curing diseases like cancer and Alzheimer’s in their top three wishes, followed by assisting people with disabilities (36%). Therapy and loneliness relief ranked near the bottom.

Fear, however, cuts much deeper and wider.

A whopping 64% of respondents cited job loss as their top concern. It was the number one fear across all 50 states, regardless of political party, running from 57% in Mississippi up to 71% in Iowa.

Interestingly, displacement anxiety rises with education. Those with graduate degrees were nearly 10 points more worried about job loss than Americans with a high school diploma or less. The knowledge workers whose day jobs overlap most directly with LLM capabilities feel the heat first.

Yet direct exposure calms nerves. Workers who use AI daily were significantly less worried about losing their jobs (54%) compared to those who never touch it (70%). Familiarity breeds practical understanding: heavy users quickly discover where the tech shines, and more importantly, where it still falls flat.

Nobody trusts AI labs to self-regulate

Americans are not losing sleep over rogue superintelligence. Instead, they fear immediate, terrestrial harms: cognitive dependency (56%), misinformation (52%), crime, and state surveillance.

And when it comes to who should manage these risks, the public verdict is brutal.

Only 15% of Americans trust AI companies to make decisions about how their models are built and deployed. Meanwhile, over 70% support government regulation, including 79% of Democrats and 68% of Republicans. The top priorities for oversight are privacy (56%), child protection (52%), and legal liability for harms (49%).

When asked how to ensure AI actually helps society, the top answer was making AI developers legally liable for the damage their tools cause.

Why it matters

For everyday workers, the biggest takeaway is psychological: the fastest cure for AI anxiety is simply using AI.

The 16-point gap between daily users and non-users proves that hands-on experience turns an existential threat into an ordinary productivity software workflow. If you have been avoiding tools like Claude or ChatGPT out of dread, sitting on the sidelines is literally making your anxiety worse.

For the tech industry, the 15% trust figure is a flashing red warning. The era of Silicon Valley asking for patience while moving fast and breaking things is dead. With overwhelming bipartisan support for government intervention and developer liability, heavy-handed legal frameworks are an inevitability.

Anthropic playing the role of neutral pollster is a calculated move. By putting public skepticism and safety demands front and center, it reinforces its own brand identity as the responsible adult in the room. But that strategy carries risks. Once strict liability laws take shape, even safety-conscious labs may find that the public's appetite for punishment will hit their own bottom line just as hard as their competitors.