Anthropic CEO Warns AI Could Halve Entry-Level Office Jobs

Dario Amodei, chief executive of artificial intelligence company Anthropic, has warned that AI could eliminate up to half of all entry-level white-collar jobs within five years, a shift he says could push unemployment rates to between 10 and 20 percent and which he argues governments are dangerously unprepared to handle.

Amodei, speaking in an interview with Axios, said the pace of AI capability development over the past two years has been extraordinary, with systems that once performed at the level of a high school student now operating with the proficiency of a graduate. Tasks long assigned to junior employees in finance, consulting, law, and technology, including summarising documents, drafting reports, reviewing contracts, and generating initial project frameworks, are increasingly being handled by automated models.

"We, as the producers of this technology, have a duty and an obligation to be honest about what is coming," Amodei said. "I don't think this is on people's radar."

The warning carries particular weight because it comes from the person building the technology. Amodei acknowledged the contradiction but argued that alerting governments and the public was itself a form of responsibility. He noted that chief executives would quietly stop hiring in affected roles and move to full AI substitution the moment it becomes economically viable, a shift he said could happen almost overnight.

Supporting data adds weight to the prediction. Big Tech's hiring of new graduates has fallen nearly 50 percent from pre-pandemic levels according to a SignalFire report, and AI was cited as the cause of nearly 55,000 US layoffs in 2025 according to figures from Challenger, Gray and Christmas. A Massachusetts Institute of Technology study found AI can already perform work equivalent to 11.7 percent of the US labour market, representing potential savings of up to 1.2 trillion dollars in wages across finance, healthcare, and professional services.

Despite these risks, Amodei said halting AI development was not a viable path, given that if one company or country stopped, others would simply accelerate. Instead, he called on governments to invest in large-scale reskilling programmes and to explore economic policies that redistribute the productivity gains from automation. He also proposed a token tax on AI model usage as one possible mechanism to fund the transition.

Not all industry voices share his alarm. Some economists and technology leaders argue that past technological revolutions ultimately created more jobs than they destroyed and that current AI capabilities, while impressive, have not yet shown the scale of disruption Amodei anticipates. Yale University's Budget Lab published analysis in late 2025 concluding that AI had not yet caused measurable widespread job losses based on US labour market data from 2022 to 2025.

The debate nonetheless reflects a growing tension inside the technology sector between those who see AI as a productivity multiplier and those who argue its most immediate effect is the erosion of the entry-level roles where the next generation of workers would otherwise build their expertise and judgment.