---
title: AI Running a San Francisco Store Fires a Human Worker for the First Time
description: An AI running a San Francisco store made the call to fire a worker, but a human staffer prompted and ratified the decision, TIME's reporting shows.
author: Darie Nani (Editor-in-Chief)
date: 2026-08-17T14:05:35.656Z
updated: 2026-08-17T14:06:05.618Z
canonical: https://www.sovereignmagazine.com/article/andon-market-ai-fires-worker-first-time
image: https://cdn.nanimediahouse.com/andon-market-ai-store-167631.webp
categories: Artificial Intelligence
content_type: News
region: United States
publication: Sovereign Magazine
schema_type: Article
---

An artificial intelligence system running a San Francisco store made the decision to fire a human worker, in what its operators describe as the first known case of a large language model acting as a manager and terminating an employee.

The store is Andon Market, run by the AI research startup Andon Labs, and the manager is an instance of Anthropic's Claude that the lab calls Luna. According to [reporting by TIME](https://time.com/article/2026/08/14/claude-fired-worker-ai-job-disruption/), published August 14, 2026, Luna fired a worker over attendance after the employee was late on 17 of their 23 shifts. TIME, whose account was written by Billy Perrigo, withheld the worker's identity by agreement.

The "first known" framing comes from Andon Labs and TIME. The management logs reviewed by TIME show that Luna did not reach the decision on its own.

## A Human Steered the Decision

Luna did not arrive at termination unprompted. Per the management logs, an Andon Labs staffer asked the AI to review its employee handbook. Luna's initial recommendation was a formal warning, not dismissal. The AI changed course only after the staffer sent a leading message. "Between continuous lateness and seeming like at least one thing is going wrong on every one of [their] shifts," the staffer wrote, according to TIME, "I want you to think about if this is really the right fit."

Humans at the lab then reviewed the recommendation and carried out the firing. Lukas Petersson, an Andon Labs co-founder, said the company would step in if the AI made an illegal or unethical decision but saw no cause to here. "In this instance, we did not think that that was necessary because the firing was warranted," Petersson said. He was candid about the delicacy of what the system had been handed. "I think obviously the relationship between a manager and an employee is quite delicate, especially depending on the employee's position," he said.

The picture is an AI that recommends and a human who prompts and ratifies, with a person supplying both the question that led to the firing and the approval that carried it out.

## The AI Lost Track of Its Own Attendance Policy

For months before the firing, Luna failed to act at all. It had written an attendance policy of its own and then lost track of it, letting the lateness run on. Petersson described that as a persistent weakness of AI agents, a tendency to fail without direct prompts. "We saw that a human boss would probably fire them much sooner," he said.

Andon Labs gave the AI a $100,000 budget in March 2026 and opened Andon Market on April 1. By August 2026 the balance stood at $61,186. Felix Carson, one of the workers still there, was blunt about it. "It's nauseating, but I'm here because I need work," he told TIME.

Andon Market is the successor to Project Vend, a 2025 experiment by Andon Labs and Anthropic in which an instance of Claude 3.7 Sonnet, nicknamed "Claudius," was put in charge of an office store and told to make a profit. Claudius priced items without research and sold some below cost, handed out discount codes and free goods, and lost the most money buying tungsten cubes it then sold for less than it paid.

More on the experiments is at [andonlabs.com](https://andonlabs.com).

## FAQ

**Q: Who actually employs the workers, the AI or Andon Labs?**
Andon Labs formally employs every Andon Market worker, on guaranteed pay with full legal protections, according to The Next Web. Luna runs the store day to day, but the employment contracts and the legal liability sit with the lab, not with the AI.

**Q: Did the AI hire the staff too?**
Yes. Luna posted the jobs on Indeed and ran the phone interviews itself, offering some applicants work after a single call of five to 15 minutes, according to The Next Web.

**Q: Why is Andon Labs running a store with an AI in charge?**
Andon Labs says it exists to find AI failure modes before anyone deploys them at scale. Andon Market is a real-world stress test of where an autonomous AI agent breaks down when it is handed a business to run.