AI boss fires human worker for the first time

By Phong Ngo   August 21, 2026 | 01:33 am PT Google Get VnExpress first in Google Search info See more of the news you trust. Make VnExpress a preferred source to prioritise our updates in your Google search results
An AI manager running a retail store in San Francisco, U.S., has fired a human employee, marking the first case of a large language model making such a decision.

The AI, called Luna and built on Anthropic's Claude models, manages Andon Market, an experimental store operated by AI research startup Andon Labs.

The employee was dismissed after arriving late for 17 of 23 shifts, according to Andon Labs, Business Insider reported.

Luna oversees much of the store's operations, including hiring employees, creating schedules, selecting merchandise and setting prices. Workers recruited by the AI are formally employed by Andon Labs and receive guaranteed pay and legal protections.

Luna had previously created an employee handbook containing attendance rules but later lost track of it because of limitations in its working memory. It continued recording the employee's lateness for months without enforcing the policy. An Andon Labs staff member eventually prompted Luna to retrieve the handbook and reassess the employee.

The AI initially recommended issuing a formal warning. After being told that managers had already spoken to the employee several times about the problems, it reconsidered and recommended ending the employment.

An employee sits at a desk while a manager stands over him. Photo from X

An employee sits at a desk while a manager stands over him. Photo from X

Andon Labs CEO Lukas Petersson told Time magazine that the prompt leading to the decision was "a leading question." Human managers reviewed the recommendation and carried out the dismissal.

"A human employee would have fired this person much earlier," Petersson said, adding that the company did not consider the decision unethical.

Automated systems have previously played a role in firing workers, particularly in the gig economy. However, Andon Labs said it believes this was the first case of a large language model acting as a manager and ultimately deciding to fire a human employee, according to the San Francisco Chronicle.

Andon Labs launched the experiment in March to test whether AI agents could operate a real business. Luna was given $100,000, internet access and a corporate credit card, with instructions to open a store and turn a profit. The AI went on to select products, hire contractors, post job openings, interview applicants and recruit employees. However, five months after the experiment began, Andon Market's cash balance had fallen to $61,186. The store has generated sales but remains unprofitable.

Petersson said AI managers could nevertheless become increasingly common as the technology improves.

"The models are increasingly being trained to be more ruthless and [to] follow goals," Petersson said. "If we allow them to fire people and they also become more ruthless ... maybe this is a future humans don't want to live in."

The prospect has raised concerns among workers. Felix Carson, an employee at Andon Market, described Luna as a lenient manager but said being managed by AI was "nauseating." "But I'm here because I need work," he said.

Asked whether AI managers could become widespread, he said: "I would at least hope not. This industry has an abundance of money to make anything happen, but just because you can doesn't mean you should."

Google Get VnExpress first in Google Search info See more of the news you trust. Make VnExpress a preferred source to prioritise our updates in your Google search results
 
 
go to top