Anthropic just published one of the most uncomfortable pieces of market research in AI history. The company surveyed 81,000 Claude users about the economics of artificial intelligence, and the headline finding is almost comically on the nose: the people who use AI the most are the most afraid it will take their jobs.

Not vaguely worried. Not theoretically concerned. Actively terrified. One software engineer told the survey: "Like anyone who has a white collar job these days I'm 100% concerned, pretty much 24/7 concerned about losing my job eventually to AI."

That quote appeared in a report published by the company that built the tool causing the fear. Let that marinate.

The Numbers Tell a Paradox

Anthropic's new Economic Index survey, conducted via its Anthropic Interviewer tool in December, found a clean correlation: for every 10 percentage point increase in a job's "observed exposure" to AI (meaning Claude is already doing a larger share of that job's tasks), perceived job threat increased by 1.3 percentage points. People in the top 25% of AI exposure mentioned displacement fears three times as often as those in the bottom 25%.

Software engineers are more worried than elementary school teachers. Not because engineers are more pessimistic by nature, but because Claude is already doing a substantial chunk of their work. They can see the trajectory. Teachers, whose jobs involve physical presence and human judgment in ways AI cannot replicate, feel less exposed because they are less exposed.

Career stage matters enormously. Early-career workers expressed far more concern about displacement than senior employees. This tracks with what we reported last week: 100,000 tech workers laid off in 2026, entry-level roles getting gutted first, seven-year Meta veterans applying at Kohl's. The junior people aren't being paranoid. They're being observant.

The Productivity Paradox Gets Worse

Here is where it gets really uncomfortable. The survey found that people in the highest-paid occupations reported the largest productivity gains from AI. The mean self-reported productivity rating was 5.1 out of 7, corresponding to "substantially more productive." One respondent said what used to take months now takes four to five days.

But the people reporting the biggest productivity boosts were also the most concerned about losing their jobs. Read that again. The workers who benefit the most from AI are the same workers who fear it the most. Because they understand better than anyone what happens when your job can be done in four days instead of four months. Eventually someone asks: do we still need five of you?

One software developer captured the squeeze perfectly: "When AI arrived, the project managers started giving harder and harder tickets and bugs to solve." The ratchet only turns one way. AI makes you more productive, your employer raises expectations, and the distance between you and replaceability shrinks.

Anthropic Published This on Purpose

This is the part that deserves scrutiny. Anthropic is a company with $30 billion in annualized revenue, an $800 billion valuation, and an IPO likely before the end of the year. Publishing a survey showing your own users are terrified of your own product is either radical transparency or calculated positioning. Probably both.

The report accompanies Anthropic's updated Economic Index, which tracks how Claude is being used across industries. The companion research paper on labor market impacts provides Anthropic's own framework for measuring displacement risk. In effect, Anthropic is building the dashboard that tells the world how much damage its own product is doing. No other AI company is doing this.

OpenAI publishes industrial policy papers proposing four-day work weeks that it won't fund. Anthropic publishes data showing its users are scared. One is corporate theater. The other is, at minimum, useful information.

The IPO Angle No One Is Talking About

Anthropic's revenue has nearly tripled from $7 billion annualized last October to over $19 billion now, with Claude holding roughly 70% of US enterprise AI spend according to Ramp data. The growth is driven by exactly the professional use cases where displacement fears are highest: coding, analysis, content creation, research.

When Anthropic files its S-1 (Goldman, JPMorgan, and Morgan Stanley are all in discussions), it will need to address the reality that its revenue growth is directly correlated with the economic anxiety of its users. The faster Claude penetrates white-collar workflows, the faster those workers worry about becoming obsolete. Revenue and resentment are on the same curve.

That is the tension at the heart of every AI IPO: you're selling automation to the people you're automating.

What Happens Next

This data arrives the same week Meta confirmed 8,000 layoffs, Microsoft offered buyouts for the first time in 51 years, and 2026 tech layoffs crossed the 100,000 mark. It arrives the day after OpenAI launched GPT-5.5, doubling API pricing while promising to replace even more human work.

Anthropic's survey gives the displacement narrative something it has lacked until now: scale. Not anecdotes, not Slate profiles of unemployed Meta workers, but 81,000 responses confirming that the people closest to AI understand its economic implications better than anyone. They are simultaneously more productive and more afraid. Both things are true, and neither cancels the other out.

The question is no longer whether AI displaces jobs. The question is what we do about it while the companies building the displacement machines are also the ones publishing the research proving it happens.