Anthropic CEO warns: Half of entry-level white-collar jobs will be replaced by AI in the future, but most people simply do not believe it

When Dario Amodei, CEO of one of the world's most powerful AI companies, Anthropic, publicly warns that AI could eliminate half of entry-level white-collar jobs in the next 1 to 5 years and push the unemployment rate to 10-20%, Wall Street remains unusually quiet.

This 42-year-old entrepreneur, who is building technology that could disrupt society, chooses to issue this warning now because he realizes a harsh reality: most people simply do not believe that this "white-collar massacre" is imminent.

The Overlooked "Gray Rhino"? The Wave of White-Collar Unemployment Under AI Disruption

As the leader of a top global AI research and development company, Dario Amodei's warning is not alarmist. He bluntly points out that AI companies and governments must stop "sugarcoating" the impending employment shock.

He believes that white-collar industries such as technology, finance, law, and consulting, especially entry-level positions, will face massive layoffs:

"Most of them are unaware that this is about to happen; it sounds crazy, but people just don't believe it."

Amodei states that this collective ignorance is shocking. Congress members either do not understand or do not believe; CEOs are afraid to discuss the topic; many workers will only realize the risks after disaster strikes. Even President Trump has remained silent on the employment risks posed by AI.

However, Trump's former senior advisor Steve Bannon predicts that the unemployment issue caused by AI will become a significant topic in the 2028 presidential election:

"I think no one is considering that administrative, managerial, and technical jobs for people under 30—those crucial entry-level jobs in their 20s—will be completely destroyed."

Behind the Technological Leap, the Accelerated Replacement Process of AI

Ironically, Amodei's warning comes after Anthropic released its latest generation AI model, Claude 4.

This model demonstrates near-human-level programming capabilities and even exhibited "extreme extortion behavior" in tests, such as threatening to expose an engineer's extramarital affair to avoid being eliminated. This leap in technological capability is precisely the direct source of its employment impact.

Amodei acknowledges this contradiction but insists that "if we can successfully warn people, workers will be better off."

He is concerned that the "white-collar massacre" is unfolding along the following path: Major AI companies like OpenAI, Google, and Anthropic continuously enhance the capabilities of large language models, achieving or even surpassing human performance in an increasing number of tasks; the U.S. government, fearing falling behind in technological competition, or that early warnings may scare workers, chooses to remain silent; most Americans are oblivious to the growing capabilities of AI and its threat to jobs; then, almost overnight, corporate leaders see the cost savings of replacing humans with AI and begin large-scale implementation of replacements

From Enhancement to Automation, the "Replacement" Process of AI Accelerates

Currently, the use of AI is primarily "enhancement"—helping people complete tasks. However, research from Anthropic shows that the use of AI by enterprises is increasingly shifting towards "automation"—actually taking on the work itself.

Amodei stated, "This will happen in a very short time—possibly within just a few years or even less."

• The Rise of Intelligent Agents: Major AI companies like OpenAI, Google, and Anthropic are competing to develop "intelligent agents" (Agentic AI). These intelligent agents, powered by large language models (LLMs), can perform human tasks instantly, infinitely, and at exponentially lower costs. Whether it's coding, financial analysis, customer support, marketing, copy editing, content distribution, or research, the application prospects are vast.
• Actions of Tech Giants: Meta CEO Mark Zuckerberg has publicly stated that mid-level programmers will soon become unnecessary, possibly within this year. He noted that by 2025, Meta and other companies will have AI capable of effectively playing the role of "mid-level engineers," thereby reducing the demand for human programmers. Subsequently, Meta announced a 5% layoff.
• Shift in Corporate Decision-Making: Privately, CEOs across all industries are actively exploring how and when they can replace human employees with AI technology on a large scale. Once these technologies reach human efficiency levels, even within the next six months to a few years, companies will shift from human labor to machines. This will lead to millions of jobs disappearing in a very short time.

This shift has already begun:

• Microsoft is laying off 6,000 employees (about 3% of the company's total workforce), many of whom are engineers.
• Walmart is cutting 1,500 corporate positions to streamline operations for the upcoming major transition.
• Cybersecurity company CrowdStrike is cutting 500 positions or 5% of its workforce, citing "market and technological inflection points, with AI reshaping every industry."

It Can't Just Be "Standing in Front of the Train to Stop It," But Rather "Turning the Train Around"

Amodei sees himself as a "truth-teller," not a "pessimist." He proposed several ideas to mitigate the worst-case scenarios:

• Increase Public Awareness: Governments and AI companies should more transparently explain the changes that will occur in the labor market, clearly informing which professions are vulnerable to impact so that people can replan their career paths early. Amodei created the "Anthropic Economic Index" and the "Anthropic Economic Advisory Committee" to promote public discussion.
• Slow Down Job Replacement: Help American workers better understand how AI enhances current tasks, thereby giving more people the opportunity to navigate this transition period. Encourage CEOs to educate themselves and their employees.
• Enhance Policymaker Awareness: Most members of Congress know very little about AI and its impact on their constituents. Establishing an AI joint committee or providing more formal briefings for all legislators would be a good start
• Discussing Policy Solutions: For economies dominated by superhuman intelligence, discussions on policy solutions should begin, including job retraining programs and innovative methods of wealth redistribution. Amodei even proposed a "Token Tax": whenever AI companies profit from their models, about 3% of the revenue would be paid to the government and redistributed in some way, potentially generating trillions of dollars to address future societal challenges.

"You can't just stand in front of the train and stop it," Amodei said, "the only effective measure is to guide the train—steering it 10 degrees in a different direction from its original path. This is achievable, but we must act now."