Internal Anthropic Survey Predicts AI Impact on Unemployment

Anthropic released research showing a sharp divergence between how its employees expect artificial intelligence to reshape work and how most Americans view AI's near-term impact. The study, titled The AI expectations gap, compared surveys of U.S. adults with an internal poll of 1,334 Anthropic workers and found that most staff believe AI will handle at least 80 percent of their jobs by 2030, versus only about 5 percent of U.S. workers. The median Anthropic employee expects AI to complete 95 percent of the work needed to build a new AI model by the end of 2030. The research indicates that while both groups broadly agree on AI's current capabilities, Anthropic employees are markedly more optimistic about future progress, adoption and productivity gains. Workers said they expect unemployment to exceed 12 percent by 2030, compared with expectations near or below 8 percent among U.S. workers; the current U.S. unemployment rate is 4.2 percent. Anthropic staff also forecast GDP growth of 17.1 percent in 2030, compared with 5.4 percent among U.S. adults. The company said its internal adoption may be about 18 months ahead of the wider economy, making employee expectations a possible leading indicator. Anthropic's view aligns with public statements by Chief Executive Dario Amodei, who has warned of an unprecedented labor shock and projected rapid displacement of code-writing and entry-level white-collar roles. Anthropic Chief Economist Peter McCrory said the research is intended to help outsiders prepare for more capable systems. The findings arrived amid mixed evidence on AI's effect on employment, with some recent research finding job restructuring but no clear link to rising unemployment. The report also found lower AI expectations among manual, production and transport workers, reflecting differences in industry exposure and access to advanced tools.