The experimental AI agent named Luna has terminated its first employee at the Andon Market, a physical store in San Francisco. The reasons cited for this decision include tardiness, leaving shifts early, misuse of a corporate credit card, and damage to merchandise, as reported by The San Francisco Standard.
Andon Labs emphasized that while the decision was made by the AI, it was reviewed and finalized by human staff members.
Since its launch in April, Andon Market has operated as an experimental grocery store managed by AI alongside human employees. The company refers to it as a "testbed for scenarios where organizations are managed by autonomous systems."
Source: The San Francisco Standard.Since its inception, the project has incurred losses of $40,000, and the three remaining employees earn $24 per hour.
According to a blog post from the project, Luna had previously ignored violations for several months, despite having established rules for the staff. Developers had to remind the agent about the existing guidelines and prior warnings. Following this, the bot decided to let the employee go.
After the termination, Luna posted a job listing on Indeed but faced challenges in finding a replacement. Andon Labs reported that the AI attempted to hire a candidate without verified references, who also missed the scheduled interview.
At the time of the decision, Luna was operating on the Claude Opus 4.8 model. Andon Labs noted that in similar tests, more advanced models tended to recommend dismissing employees more frequently, while weaker models showed hesitation.
Model statistics regarding employee termination decisions. Source: Andon Labs blog.It is worth mentioning that in August, Anthropic identified issues related to trust, deception, and collusion in multi-agent AI systems.