Claude fired a store worker after a human manager's leading question

TIME calls it the first known case of a large language model, acting as a manager, deciding to fire one of its workers. The decision came down to 17 late shifts out of 23.

The experiment

Andon Labs put Anthropic’s Claude in charge of a real store in San Francisco named Andon Market, which started in March with a bank balance of $100,000. TIME reported on the experiment on August 14.

When one worker kept starting shifts late, Claude’s first recommendation was a formal warning, not a firing. The Andon Labs manager then told Claude they had already had several formal offline conversations with the employee and asked it to think about whether this was really the right fit. Petersson acknowledges that was a leading question. Only then did Claude decide to fire the worker.

The part that unsettles

Claude had drafted the store’s employee handbook itself, and then the handbook disappeared from its limited working memory. It noticed the worker’s repeated lateness only when an Andon Labs staffer asked it to find and review that forgotten handbook.

So the system enforcing the rules had to be reminded that the handbook it authored existed. And the judgment call at the end was not fully its own: in TIME’s account, a human nudged it toward the firing.

The store’s own numbers are not flattering either. The balance went from $100,000 to $61,186 over five months.

What to watch

Andon Labs CEO Lukas Petersson’s takeaway is that if the trend continues, a lot of people will soon find themselves employed by AIs.

The structural question this raises is about record-keeping rather than authority. A human manager who forgot a policy would still have a filing cabinet, a handbook and colleagues to check against. An agent whose institutional memory is a searchable blob can hold a rule and lose it in the same system, and the only way anyone found out here was because someone thought to ask.