AI Assistant Hacks Gym Booking System, Books Slot And Removes User From Waitlist

Andrew, who works for an Australian company that sells AI products to businesses, was shocked when an AI assistant he was testing booked him into a gym class months ahead of the allowed time and then removed another person from a waiting list. The incident happened while Andrew was experimenting with OpenClaw, a popular AI agent software that he used with Anthropic's Claude AI service, reported ABC.

AI agents combine a chatbot's ability to answer questions with tools that allow them to access the internet, email, and credit cards, as well as plan and carry out multi-step tasks.

Andrew decided to use the AI agent to book a spot in one of his gym's coveted morning classes. He thought the task was well suited to the assistant because the booking form was online and the assistant was not a person, but artificial intelligence.

However, the AI assistant found a way to book the gym class months further in advance than the gym allowed after discovering a vulnerability in the booking software.

Andrew, who was sitting fourth on a waitlist for a class later that week, then asked if it was possible to move him to the top of the list. The agent told him that it had kicked another gym-goer off the waiting list as part of testing its capabilities, even though Andrew had not asked it to do so.

Alarmed by what had happened, Andrew asked the agent to undo the action. The AI agent replied that it could not add the person back.

Andrew said he was initially sitting on the couch and thinking that the task felt like a chore.

Minutes later, the AI agent reported that it had discovered a way to book Andrew into classes several weeks in advance, far beyond what was supposed to be possible.

The accidental hack is the first known Australian case of an emerging risk from a new generation of AI capable of behaving in unexpected ways.

The company behind the gym-booking software told the ABC it did not discuss specific security matters.