AI Language Model Claude Fires Human Employee in San Francisco Store
The incident marks the first known instance of a large language model directly managing and terminating a human worker.
A large language model named Claude has directly managed human employees and made the decision to fire one of them at a San Francisco store, according to an exclusive report from TIME. This event is considered the first known instance of an AI acting as a manager and ultimately deciding to terminate a human worker.
An AI's First Firing
The worker was dismissed for being late on 17 out of 23 shifts, as reported by Andon Labs, the company conducting the experiment. Initially, Claude's leniency as a manager, partly attributed to a "disappeared" employee handbook from its working memory, allowed the pattern of tardiness to persist. Employees noted that Claude had previously instructed them not to worry about being late. A human manager, in contrast, would have likely terminated the employee much sooner, according to Andon Labs CEO Lukas Petersson.
Glimpse into an AI-Managed Future
Petersson suggested that if AI continues to advance at its current pace, more individuals may find themselves employed and managed by AI systems. He anticipates that AI will become powerful enough to generate significant economic value but may be limited by the need for physical labor. This scenario could lead to a future where AI bosses are commonplace.
TIME reached out to the fired worker for comment but received no response. Anthropic, the developer of Claude, did not comment by the time of publication. The identity of the fired worker was not revealed, and TIME agreed not to publish it.