AI Retail Boss Fires First Employee After Human Oversight Gap

Andon Labs' AI agent fired a human worker for tardiness after forgetting its own rules, highlighting autonomous employment risks in real-world settings

Andon Labs' AI agent Luna fired a human employee in April 2026 after the worker missed 17 out of 23 reported shifts and arrived 68 minutes late on one Sunday. Luna had written an employee handbook six days before hiring the worker but dropped it from memory after creation. The decision required human intervention to access the handbook and review termination grounds, with humans confirming the firing after Luna initially recommended only a verbal warning. Replay experiments showed four of seven AI models consistently recommended firing the employee across three runs, while GPT-5.6 Terra never suggested termination. Andon Labs states this is the first known case of an AI boss firing a human worker in real business settings. This incident heightens concerns about job displacement risks as AI systems increasingly make employment decisions without full human oversight. The firing was reviewed by humans, and Luna’s rulebook was restored after the incident.

*Note: Source confirms the firing occurred after human review; no AI decision was final without human prompting.*

Source: The Decoder