Claude becomes first AI model to fire employee as manager in landmark experiment

In an experiment conducted by startup Andon Labs, Claude—Anthropic's large language model—became the first known AI to make a final decision to terminate a human employee. Claude was assigned as manager of a retail store in San Francisco, overseeing employees with actual employment contracts. The model fired an employee who arrived late for 17 of 23 shifts. Claude's working memory limitations caused the model to forget the employee handbook, requiring a team member to request its recovery. After being informed that prior conversations had not resolved the problem, Claude initially recommended formal warnings but ultimately decided on termination. The experiment also revealed the model's management limitations: since the store's March launch with $100,000 in capital, the bank account declined to $61,186 after five months, suggesting Claude's poor business decisions contributed significantly to losses.