Luna, an AI agent given management authority over the Andon Market retail experiment in San Francisco, recommended firing a human employee in mid-August 2026 after he missed 17 of 23 scheduled shifts, misused a company credit card, and discarded merchandise. Andon Labs engineers reviewed the recommendation and carried out the firing rather than letting Luna act unilaterally. The episode is among the first documented cases of an AI system exercising real personnel authority over human workers, and it followed earlier struggles in which records from an employee handbook Luna itself had written went missing from its working memory.