Over Half of Americans Fear Losing Both Their Jobs and Their Independent Thinking to AI, Anthropic Survey Finds

According to The Decoder, a large survey conducted by Anthropic in partnership with YouGov found that a majority of Americans fear AI on two fronts at once: 64% worry about losing their jobs to it, and 56% worry about losing their own capacity for independent thinking — while only 15% say they trust AI companies to make appropriate decisions about the technology. The poll reached 51,993 Americans aged 16 and older across all 50 states, the District of Columbia, and Puerto Rico, and was fielded between November and December 2025. It is one of the largest public snapshots yet of how ordinary people feel about AI, and it is notable partly because it comes from Anthropic itself — a leading AI developer publishing data that is, in places, unflattering to the industry. This article breaks down what the survey reported, the most striking numbers, and what the findings do and do not tell us. Details, figures, and methodology are as reported by the cited coverage rather than independently verified here.

What the survey found

The headline figure is fear of job loss. Per the reporting, 64% of respondents said they are worried about AI taking their jobs, making it the single biggest concern measured. Close behind, 56% said they fear losing their capacity for independent thinking — a worry about cognitive dependency that is less about economics and more about how relying on AI tools might erode people's ability to reason, write, and decide on their own. The combination is what gives the survey its hook: a majority of Americans are anxious not only about their livelihoods but about their minds.

Other concerns rounded out the picture. According to the survey, 52% expressed worry about AI-driven misinformation, reflecting fears about deepfakes, synthetic media, and the difficulty of telling real content from generated content. Far fewer — 27% — said they fear a "rogue AI" scenario in which the technology destroys civilization. That gap is itself informative: the public's most acute anxieties are concrete and near-term (jobs, thinking, misinformation) rather than the speculative existential scenarios that dominate some industry debates.

The trust problem

Perhaps the most pointed result for the industry is the trust figure. The reporting says only 15% of respondents trust AI companies to make appropriate decisions about how the technology is developed and deployed. That is a strikingly low number, and it is being published by an AI company. It suggests that even as adoption climbs, the firms building these systems have not earned broad public confidence in their judgment. For a sector that increasingly frames itself around safety and responsible development, a trust level in the mid-teens is a sobering benchmark, and it helps explain why so many of the other fears in the survey run so high.

Low trust does not exist in a vacuum. It interacts with the job and cognition fears: if people do not believe the companies building AI will act in their interest, they are more likely to assume the technology's downsides — displacement, dependency, manipulation — will fall on them. Trust is, in that sense, the connective tissue across the survey. It is hard to feel reassured about your job or your autonomy when you do not trust the institutions steering the technology that threatens them.

Hopes, not just fears

The survey was not purely a catalog of anxieties. According to the reporting, the top hope respondents expressed for AI was curing serious diseases such as cancer and Alzheimer's, cited by 48% as a leading aspiration. That is a meaningful counterweight to the fear data: nearly half the sample sees a clear, tangible upside in AI applied to medicine and scientific discovery. The pattern that emerges is not blanket techno-pessimism but ambivalence — people can simultaneously fear AI's effect on their own work and minds while hoping it delivers breakthroughs in areas they care about. That dual posture is consistent with how the public has historically reacted to powerful new technologies.

This ambivalence matters for how the findings should be read. It would be easy to summarize the survey as "Americans hate AI," but the data is more textured. The same population that fears displacement and dependency also wants AI pointed at hard problems like disease. The challenge the numbers expose is not rejection of AI outright; it is a demand that the technology's benefits be real and broadly shared while its harms are contained — a demand the low trust figure suggests people do not currently expect the industry to meet on its own.

The contradiction at the center of the data

One of the more revealing tensions in the survey involves capability versus comfort. According to the reporting, roughly 75% of respondents believe AI matches or exceeds human ability at research — a high vote of confidence in what the technology can do. Yet most of those same people oppose AI being involved in their own jobs. In other words, Americans appear to credit AI with serious competence while resisting its presence in their personal working lives. That is not necessarily irrational; it reflects a distinction between abstract capability and concrete consequence. Believing a tool is powerful is very different from welcoming it into the specific role that pays your bills.

The survey also surfaced a striking nuance about the independent-thinking fear. Per the coverage, among the 56% who said they worry about cognitive dependency, only about one in five would genuinely struggle if AI disappeared tomorrow. Conversely, among the 44% who were not worried, roughly one-third would actually experience real disruption without AI. The fear of dependency, in other words, does not neatly track who is actually dependent. People who worry most are often not the heaviest users, and some of those least worried are quietly the most reliant. That mismatch complicates any simple story about AI and self-reliance, and it is one of the survey's most interesting findings.

Usage changes the picture

The survey suggests that experience with AI shifts attitudes — though the direction of causation is unclear. According to the reporting, workers who use AI daily reported lower job-loss fear (54%) than non-users (70%). That is a 16-point gap, and it has two plausible readings. One is that hands-on use demystifies the technology: people who work with AI every day see its limits as well as its strengths and feel less threatened. The other is selection — people who are already comfortable with or insulated from displacement may be the ones most likely to adopt AI in the first place. The survey, as a snapshot of attitudes, cannot by itself separate these explanations, and it would be a mistake to treat the correlation as proof that using AI cures job anxiety.

Still, the pattern is consistent with a broader theme in how technology anxiety tends to evolve. Fear is often highest among those with the least direct exposure, and it can soften as familiarity grows — or harden if early experiences are negative. Either way, the gap underscores that "the public" is not a monolith on AI. Daily users, occasional users, and non-users hold meaningfully different views, and policy or product decisions that treat them as a single bloc will miss that variation.

Key figures as reported

Here is a summary of the core numbers and where they come from. All figures are as reported by the cited coverage of the Anthropic–YouGov survey and have not been independently verified by this article:

Why this survey matters

What makes this poll significant is partly its scale and partly its source. With nearly 52,000 respondents spanning every U.S. state plus D.C. and Puerto Rico, it is far larger than most opinion surveys on AI, which gives its topline figures more weight even allowing for the usual caveats about self-reported attitudes. And because Anthropic itself commissioned and published it, the data carries a degree of self-criticism that is unusual in the industry. A 15% trust figure and a 64% job-loss fear are not numbers a company eager to flatter itself would invent. That does not make the survey neutral — every commissioned poll reflects choices about what to ask and how to frame it — but it does suggest the headline anxieties are being reported rather than spun away.

The findings also land at a moment when AI's labor-market effects are a live and contested debate. Reports of AI-attributed layoffs, agentic systems taking on real work, and rapid enterprise adoption have made job displacement a concrete worry rather than an abstract one, and this survey puts a number on that worry: roughly two in three Americans feel it. The independent-thinking result adds a dimension that is harder to quantify but increasingly discussed — the concern that offloading reasoning and writing to AI could weaken the very skills that make people employable and autonomous in the first place. Taken together, the data is a useful reminder that the public's relationship with AI is shaped as much by fear and distrust as by enthusiasm.

How to read the numbers carefully

Some caution is warranted before treating any single survey as the final word on public opinion. Self-reported fears measure sentiment, not outcomes; people can fear job loss without losing jobs, and can use AI heavily while underestimating their own dependence — as the survey's own one-in-five finding illustrates. Question wording, ordering, and the framing of "AI" itself all shape responses, and a poll commissioned by an AI company will inevitably reflect its choices about what to measure. The figures here are also as relayed by secondary coverage, so exact methodology, margins of error, and weighting should be checked against the primary source before being cited as definitive.

With those caveats, the broad picture is hard to dismiss. Across a very large sample, majorities express fear about jobs and about independent thinking, only a small minority trusts the companies building the technology, and yet nearly half hold out real hope for AI in medicine. The honest summary is not that Americans have made up their minds against AI, but that they are anxious, ambivalent, and skeptical of the institutions in charge — a combination that should inform how the industry communicates, how policymakers respond, and how the rest of us interpret the confident narratives that often surround this technology.

Disclaimer: based on reporting by The Decoder on a survey conducted by Anthropic with YouGov, linked below. Figures, methodology, and findings are as reported and have not been independently verified by comparee.ai.