Anthropic: Employees Predict Over 12% Unemployment by 2030

A survey conducted among 1,334 employees of Anthropic presents forecasts that are much more pronounced than those of other American workers. The expectations encompass both a rise in unemployment and a surge in growth, as Dario Amodei, the CEO, proposes rapid scenarios for the automation of coding and office jobs.

A Fragile Job Market and Boosted GDP, According to Respondents

Most of the Anthropic employees surveyed believe that the unemployment rate will exceed 12% by 2030. In contrast, the majority of American workers expect it to be around 8% or less, compared to the current rate of 4.2%. The aggregated responses also provide growth projections: the median estimate for American adults for 2030 is a GDP growth of 5.4%, while Anthropic employees estimate it at 17.1%. For reference, GDP grew by 2.2% last year.

Employer Statements Align with High Expectations

In March 2025, Dario Amodei stated that he expects AI to write 90% of the code within six months, and "essentially all the code" within a year. In May, he added that it could eliminate half of entry-level office jobs within five years, with unemployment rates at 10% or 20%. In a subsequent essay, he anticipates an "unprecedented" short-term shock to the labor market and strong growth, arguing that AI would serve as a general substitute for human labor rather than for specific jobs. These positions make him a particularly influential voice on these topics.

Tasks Taken Over by AI: Massive Gaps in Expectations

Most Anthropic employees believe that AI will take over at least 80% of their tasks by 2030, while only about 5% of American workers express the same expectation. The average respondent at Anthropic estimates that 95% of the work required to create a new AI model will be performed by systems by the end of 2030. These expectations are accompanied by fears of professional obsolescence that are not widely shared by other workers.

Current Usage: Agreement on Capabilities, Divergence on Speed

Surveys assessed the current use of AI and upcoming changes. American adults and Anthropic staff largely converge on the current capabilities of the tools. However, they diverge on the speed of improvement, the extent of use, and the expected productivity gains. Among the explanations offered are higher tool budgets at Anthropic and access to unpublished models, while American workers are more prevalent in manual and service jobs. Respondents in manual, production, and transport jobs thus predict that AI will take over the smallest share of their tasks by 2030, compared to Anthropic employees, those in the tech sector, and in business and finance.

Why Publish These Internal Data, According to Anthropic

The initiative is based on 1,334 employees and compares their responses to those of American adults in a study titled "The AI Expectations Gap." Anthropic researchers describe a consensus on current capabilities and a gap regarding future trajectories. Peter McCrory, chief economist, argues that these results can help prepare for more advanced systems, as the company claims to be about a year and a half ahead of the economy in terms of adoption, potentially making it a leading indicator. Evidence on the impact of AI on employment remains mixed; a Yale team recently observed job changes unrelated to fluctuations in unemployment. Finally, the responses were aggregated to derive macroeconomic scenarios.