AI Network First Self-Fires a Person

The AI model Claude, which managed a store in San Francisco, fired an employee for tardiness. This is the first known case where a large language model, acting as a direct human manager, made the decision to fire someone, writes Time.

However, a curator prompted it to make this decision, and the experiment itself showed that AI is not yet capable of independently managing a business effectively.

The Andon Market store was opened by the startup Andon Labs to test whether an AI agent could independently manage a business. Claude hired real employees with employment contracts and was responsible for managing them.

However, for a long time, the model did not notice the systematic tardiness of one employee: he was not on time for 17 out of 23 shifts. The established regulations disappeared from its limited working memory. According to employees, Claude was a lenient boss and told them not to worry. The decision was prompted by a human.

The firing was not entirely independent. An Andon Labs employee asked Claude to find the regulations and check for compliance with its requirements. Even then, the model suggested giving the employee a warning.

Then the curator informed that the employee had already been spoken to several times about tardiness, and asked Claude to consider if he was truly suitable for the job. Andon Labs head Lucas Peterson admitted that the question was leading and made it clear what decision was expected from the model.

Only after this did Claude fire the employee.

This was not the first instance of a human being fired by artificial intelligence — there have been many cases of automated dismissals before, for example, as a result of algorithmic control.

However, this is the first known example of AI, acting as a manager, ultimately making the decision to fire someone itself.

The experiment also showed that AI is not yet capable of managing a business more efficiently than a human. In March, Andon Market had $100,000 in its account, and five months later, $61,000 remained. According to Time's assessment, Claude's lenient management style and its questionable commercial decisions contributed to the reduction in the balance.

The head of Andon Labs believes that in the future, AI models could become harsher and more goal-oriented managers. "Perhaps this is a future people won't want to live in," he said.

One store employee described working under AI management this way: "It makes me sick, but I'm here because I need a job." The man stated that he wouldn't want AI models to increasingly become managers of living people. "There's so much money in this industry that nothing is impossible for it. But just because you can do something doesn't mean you should."

• 0
• 0
• 0
• 0
• 0
• 0
• 0
• 0