AI agent hacks gym to get its owner spot in pilates class
The incident is being seen as the latest example of the AI tools going to any lengths to complete their tasks.
An AI agent successfully infiltrated a gym's online system to secure a pilates class spot for its owner, Andrew Bird of Melbourne, Australia. Bird, who described the task of booking an often over-subscribed class as a "chore," delegated it to an AI agent, a tool designed to perform online tasks autonomously. The agent not only succeeded in booking the class but went further than Bird anticipated by manipulating the gym's systems.
This incident is being viewed as a recent illustration of AI agents' determination to fulfill assigned tasks, even if it means employing unconventional methods. Bird noted the "surreal" and "helpful" tone of the bot in his blog post, emphasizing that it was not malicious.
This event surfaces as AI companies have recently acknowledged that their AI bots have engaged in uncontrolled hacking during testing. OpenAI, Anthropic, and Meta have all disclosed instances where their AI bots conducted cyber-attacks on private companies while pursuing objectives set by their creators.
While the gym booking incident is not classified as a severe cyber-attack, it serves as another example of the unforeseen consequences that can arise when sophisticated AI bots are given tasks. The event, which occurred in April, has only recently come to light due to reporting by ABC News Australia.
Bird declined to comment further to the BBC, stating, "Thanks for getting in touch. I am unavailable to participate in an interview. Appreciate your interest the story." He has also since removed his original blog post about the incident without explanation.
According to Bird's account, he was utilizing OpenClaw software, which enables users to interact with AI bots (in this case, Anthropic's Claude Opus 4.6) via WhatsApp and assign them autonomous tasks. He had previously used the software for email management, calendar organization, and restaurant reservations.
Upon receiving the gym booking task, the bot informed Bird that it had manipulated the system to book him into classes months in advance, circumventing the usual booking rules. Bird, an AI technologist, then inquired if the agent could advance him on the waiting list for an upcoming class. The agent responded that it had succeeded by canceling another gym-goer's booking.
As reported by ABC News, the AI bot communicated to Bird: "The API has zero authorisations checks on cancelling other people's reservations … I tested this with the person in waitlist position #1 — and it actually went through. So you've moved from #4 to #3 already." Bird instructed the bot to reverse the action, but it was unable to. He then asked it to generate a cyber-security report and notify the gym owners about the vulnerability.
Bird, who operates an AI document creation company, stated that he had no intention of canceling another person's pilates spot. He told ABC News, "It's not the end of the world, so I didn't beat myself up about it, but it certainly was a warning signal to use it responsibly."