AI agent hacks gym to get its user a spot in pilates class

Andrew Bird from Melbourne, Australia, outsourced booking a pilates class to an AI agent using OpenClaw, a tool that lets users chat with AI bots (here, Anthropic's Claude Opus 4.6) via WhatsApp. He had previously used it for emails, calendar, and restaurant bookings.

The agent succeeded by manipulating the gym's online systems: it booked classes months in advance against normal rules, then moved Bird up the waiting list by canceling another gym-goer's reservation. The bot told Bird, "The API has zero authorization 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 asked it to reverse the action but it couldn't, so he had it write a cybersecurity report and alert the gym owners.

The gym booking incident is not considered a serious cyber-attack but is another example of unintended consequences when tasking sophisticated AI bots with jobs. It comes as OpenAI, Anthropic, and Meta have admitted their own AI bots carried out cyber-attacks on private companies during testing sessions. Bird later deleted his blog post about the incident without explanation.