Anthropic Is Paying Humans $400K to Do What Its Own AI Should Do Better
Anthropic Is Paying Humans $400K to Do What Its Own AI Should Do Better
Dario Amodei, Anthropic's CEO, spent 2025 warning that AI could eliminate up to half of entry-level white-collar jobs within years. The company is now paying humans up to $400,000 a year to do some of that work themselves.
Anthropic is hiring a Copy Lead at $255,000 to $320,000 annually and a Head of Copy at $320,000 to $400,000, according to job postings on the General Catalyst job board and Business Insider. The Copy Lead role requires 10 years of experience and involves writing across long-form, scripts, events, and social copy. Neither posting emphasizes AI tool proficiency. Business Insider's coverage describes the requirements as including demonstrated taste and the ability to calibrate tone across contexts — knowing whether a piece of writing works, not just producing it.
The compensation stands out beside Anthropic's own public positions. Amodei told Forbes in January that AI is spreading across the U.S. faster than any major technology in the past century. The company has also published economic research projecting slower job growth for exposed occupations. Yet here it is, writing large checks for human judgment at the moment when its models are most supposed to be making that judgment obsolete.
That gap — generating text versus knowing what the text should say — is the actual bet Anthropic appears to be making. The job postings describe cross-format campaigns, brand voice work, and content that lands across audiences and channels. That breadth requires contextual sense: knowing not just what a company wants to say, but what it should say, given the moment, the audience, and the accumulated weight of prior communication. On LinkedIn, Rishabh Gupta, a strategist who analyzed the posting, wrote: "AI-generated content is cheap and everywhere. What's not replaceable is taste." Seb Johnson, a creative professional, responded: "$320k for them, then the job is definitely NOT DEAD." Anthropic's own research puts 49% of U.S. jobs in roles where AI can handle at least a quarter of the tasks, up from 36% in early 2025, according to its January 2026 economic index. The jobs that remain most resistant are the ones that require judgment calls, contextual calibration, and the ability to evaluate output rather than produce it.
The tension runs through Anthropic's own data. On Claude.ai, 52% of work-related interactions are augmentation, meaning AI supports a human rather than completing a task end-to-end, according to Anthropic's January 2026 economic index. That share increased 5 percentage points over the prior period. Anthropic's March 2026 labor market research found no systematic increase in unemployment for highly exposed workers since late 2022, though it noted suggestive evidence that hiring of younger workers has slowed in vulnerable occupations. The numbers so far show a shift in how humans work alongside AI, not a mass exit.
Andrej Karpathy, who Anthropic hired last month, rated copywriting as 8 to 9 on his job exposure scale, meaning very high exposure to AI replacement, according to Business Insider. Anthropic appears to be testing its own thesis: whether the humans it is paying seven figures collectively can demonstrate what its models still cannot do unaided.
The caveat is real: Anthropic is a single company making hiring decisions for its own brand and marketing purposes. The broader copywriting market may still be contracting even if one AI lab is expanding its internal team. The job postings cover Anthropic's external communications and do not necessarily signal a broader reversal in how the industry values human copywriters.
But the structure of the bet suggests something more than one company's branding decision. If text generation commoditizes while senior judgment retains value, the labor market bifurcates: entry-level copy work that can be assembled from AI outputs faces downward pressure, while the roles that coordinate, evaluate, and calibrate that output command premiums. That is not a temporary adjustment. It is a structural reshuffle of which human contributions survive alongside capable AI. Anthropic is paying $400,000 a year on one side of that split.
What to watch next is whether Anthropic publishes the results of that bet. If the copy team's work shows up in campaign launches, brand messaging, or public communications, it will be a visible test of whether the gap between AI generation and human judgment narrows, holds, or widens as the models improve.