The experimental AI agent Luna has fired its first employee at the Andon Market offline store in San Francisco. The reasons cited were tardiness, early departures from shifts, misuse of the corporate card, and damage to goods, according to The San Francisco Standard.
Andon Labs emphasized that while the decision was prepared by artificial intelligence, it was reviewed and formalized by humans.
Andon Market has been operating since April as an experimental grocery store managed by AI with a human staff. The company describes it as a “testbed for a scenario where organizations are run by autonomous systems.”
Since its opening, the project has lost $40,000, and the three remaining employees earn $24 per hour.
According to a blog post from the project, Luna did not respond to violations for several months, despite having set the rules for the staff. Developers had to remind the agent of the existing protocols and previous warnings. Following this, the bot decided to terminate the employee.
After the dismissal, Luna posted a job opening on Indeed but encountered issues in finding a replacement. Andon Labs stated that the AI attempted to hire a candidate without verified references, who missed the scheduled interview.
At the time of the decision, Luna was operating on the Claude Opus 4.8 model. Andon Labs added that in similar tests, stronger models more frequently recommended dismissing the employee, while weaker ones hesitated.
In August, Anthropic identified issues with trust, deception, and collusion in multi-agent AI systems.
