AI Has Yet to Trigger Broad Job Losses, Anthropic Says

Since ChatGPT’s late-2022 debut, rapid advances in generative AI have fueled warnings that software could displace large swaths of white-collar labor. Anthropic sought to test that thesis with an “observed exposure” measure combining Claude usage, O*NET task data and U.S. labor statistics. The framework distinguishes what large language models can theoretically do from tasks they are actually automating at work, offering a clearer gauge of whether AI is augmenting employees or eliminating demand for them.

Anthropic head of economics Peter McCrory said on July 22, 2026, that AI had not materially increased unemployment, including among workers most exposed to automation, and was unlikely to do so over at least the next year. Anthropic’s March 5 study found no systematic rise in unemployment in highly exposed occupations since late 2022. It did, however, identify a tentative warning: the job-finding rate for workers aged 22 to 25 entering exposed occupations was 14% below its 2022 level, while no comparable decline appeared for workers over 25.