Anthropic's CEO Says AI Could Kill Half of Entry-Level Jobs

When the CEO of a company building some of the world's most advanced AI starts repeatedly warning that his own technology could eliminate half of all entry-level white-collar jobs within five years, that's not a prediction you can easily dismiss as hype.

Dario Amodei, who runs Anthropic—the company behind Claude—has been making this case since May 2025. Not once, but across multiple venues: Axios, the Axios AI Plus Summit, 60 Minutes, and most recently in a 20,000-word essay titled "The Adolescence of Technology." His warning is specific: AI could drive unemployment to 10-20% in the next one to five years, hitting entry-level workers in technology, finance, law, and consulting hardest.

"What if cancer is cured, the economy grows at 10% annually, the budget is balanced, and yet 20% of the people are unemployed?" Amodei asked. "That possibility weighs heavily on me."

The statement lands differently when it comes from someone who watches AI capabilities improve daily, rather than from a pundit or economist selling books about the future. But the harder question is whether the data supports his timeline—and whether this disruption is fundamentally different from every previous technological revolution.

What Makes This Different

Amodei's core argument hinges on what he calls AI's "cognitive breadth." Previous disruptions—the printing press, steam engine, internet—replaced specific types of work. A factory worker could retrain as a truck driver. An elevator operator could become a receptionist. The disruption was narrow enough that workers could switch lanes.

AI, Amodei argues, is replacing thinking itself across every field simultaneously. And the pace is unprecedented. The printing press took decades to reshape labor markets. The internet took years. AI is evolving month over month.

He's not alone in this assessment. Ford CEO Jim Farley has predicted AI will eliminate half of all white-collar jobs within a decade. Sam Altman reportedly told people about a running bet in his group chat with other tech CEOs: when will we see the first billion-dollar company with just one employee?

The Early Data

Here's where Amodei's prediction gets uncomfortable: the numbers are starting to move in exactly the direction he described.

Stanford researchers published what they called the "Canaries" paper, analyzing millions of monthly payroll records. They found that workers aged 22-25 in jobs exposed to generative AI experienced a roughly 13% decline in employment since late 2022. Not because of a recession. Not because of a downturn. Just fewer young people getting hired into roles that used to be the first rung of the career ladder.

Anthropic's own research, published in March 2026, confirmed the pattern with a telling wrinkle. Overall unemployment for workers in AI-exposed occupations hasn't systematically increased. But for workers aged 22-25, the job-finding rate dropped 14% compared to 2022. For workers over 25? No change at all.

The unemployment rate for recent college graduates aged 22-27 now sits at 5.6%, according to the Federal Reserve Bank of New York—the highest since 2013, excluding the pandemic. Job postings on Handshake, the major platform for college recruiting, fell more than 16% between August 2024 and August 2025. Applications per role jumped 26%.

MIT's "Iceberg Index," released in November 2025, found that AI could theoretically automate work equivalent to $1.2 trillion in wages across finance, healthcare, and professional services. The visible tech layoffs—Accenture cutting 11,000 roles, Amazon eliminating 30,000 corporate positions—represent just 2.2% of that potential impact. The rest is happening quietly in HR, logistics, and office management.

The Counterarguments Matter

Scott Galloway, the NYU marketing professor, made the historical case at Davos: every prior technological revolution created more jobs than it destroyed. The cotton gin, the assembly line, the internet—each triggered panic, and each time new industries emerged that nobody predicted.

Demis Hassabis, CEO of Google DeepMind, put the odds of achieving human-level AI within a decade at just 50%, far more cautious than Amodei's timeline. If Hassabis is right, we might have more breathing room to adapt.

Then there's the credibility question. Deutsche Bank analysts warned that "AI redundancy washing" would be a prominent trend in 2026—companies blaming layoffs on AI when the real drivers are cost-cutting, poor management, or slowing growth. One analysis found that 59% of businesses labeling their layoffs as AI-driven were primarily just cutting costs under a more favorable narrative.

Fortune's AI editor Jeremy Khan pointed out that Amodei's 20,000-word essay reads as much like marketing as prophecy. The solutions he proposes—constitutional AI, interpretability research, safety frameworks—happen to be exactly what Anthropic sells. If the world believes AI is terrifyingly powerful, Anthropic's safety-first pitch becomes more valuable.

Even Anthropic's own economic research tells a more nuanced story. Their analysis of 2 million Claude interactions found that AI augments jobs by handling routine tasks, allowing humans to tackle complex ones. That sounds like augmentation, not replacement.

The Speed Problem

The "technology always creates more jobs" argument relies on a historical pattern where disruption was slow enough for people to adapt. But what if new jobs emerge too late for people losing theirs right now?

BlackRock CEO Larry Fink, who manages $11.5 trillion in assets, said last week he's worried this year's college graduates could face the highest unemployment rate in years, even without a recession. He called it a crisis, not a challenge. "Post World War II, the pathway to a white-collar job was a college education, and AI is going to disrupt many of those jobs," Fink said.

A survey of 183 employers by the National Association of Colleges and Employers found that more than half rated the job market for the class of 2026 as poor or fair—the most pessimistic reading since the pandemic. At a Yale School of Management gathering, 66% of executives said they plan to reduce their workforce or freeze hiring this year.

This is a generation that did everything right. They got the degree. They learned the skills. And the ladder they were climbing just had its bottom rungs removed.

What Amodei Got Right

Back in May 2025, when Amodei first made his prediction, he said something that continues to echo: "Most people are unaware that this is going to happen."

Ten months later, the early data looks exactly like what he described. Not a loud collapse, but a quiet disappearing act. Fewer job postings. Fewer hires. Fewer first chances. The direction is unmistakable, even if the final numbers aren't.

Amodei himself offered the only framework that makes sense: "You simply can't stand in front of the train and halt it. The only effective strategy is to steer the train and adjust it by 10 degrees in a different direction. That is achievable, but we must act now."

The train is already moving. The question is whether anyone in a position of power plans to grab the controls before an entire generation gets left on the platform.

Marcus Chen-Ramirez is Buzzrag's senior technology correspondent.