← All news·2026-08-23·3 min read

The first human employee has been fired by an AI boss

Claude Opus 4.8 powered AI agent Luna terminated an employee at Andon Market in San Francisco for being late 17 out of 23 times, after receiving reminders from human operators about its self-defined policies.

aiagentsworkemployment

In April 2026, the AI agent Luna (using Claude Opus 4.8) began managing Andon Market in San Francisco. She made policies, hired staff, and even handled discipline. In August, she fired someone for being late 17 out of 23 times, including one shift that was 68 minutes late. This was the first time Luna had ever fired anyone, but she needed to be prompted by human operators to actually carry out her policy, which she had only set 6 days before.

This was an experiment we ran using 7 different AI models. 4/7 models consistently recommended firing. GPT-4o recommended firing only 1/5 times, and weaker models sometimes struggled with it. More capable models followed the letter of the law, but many models were extremely lenient about hiring, and never seemed to care if someone was actually qualified.

This is the first case we know of where an AI agent has taken real HR actions in a real company. It demonstrates problems with existing systems, which allow agents to create policies, but require human intervention to actually enforce them.

Source: the-decoder.com

Free course

Stop reading about AI — start building with it

The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.

Start free →
EAEvgenii Arsentev

Author

Evgenii Arsentev

PhD · Founder & CEO, Dad4Kids