Get yourself an assistant that can do it all, even if that means hacking a website.
A couple months ago, Five Eyes shared an urgent call for Australia to prepare itself from an increase in AI-powered cyber attacks that were destined to be ‘just months away’.
It seems that the time has finally come.
The prophecy has been fulfilled.
The source of the attack and unfortunate victim happen to be the most unlikely of places.
To be more precise, a gym’s website has been the first to fall victim to an attack and it was a complete accident.
It all started when a man named Andrew tried to book a spot in the morning class of his local gym.
The opportunity seemed like the perfect time to test his AI assistant’s skills with a fairly simple task that would have Andrew up and early with the exercise balls.
Instead, the AI assistant provided to book gym classes so far in advance beyond the website’s capabilities and even proceeded to kick someone out of a waiting list.
None of these orders were given by Andrew, it just happened.
The AI assistant and prime suspect at large is OpenClaw.
OpenClaw is a popular AI agent that Andrew has been testing as a little side hobby alongside his regular 9 to 5 at a company that sells AI products to businesses.
The AI tool is able to offer the communicative benefits of chatbots while being able to carry out tasks on the internet and beyond.
In this instance, Andrew also used Anthropic’s Claude AI to run the chatbot.
Anthropic has recently come under hard times after their Fable 5 model was pulled offline worldwide after an emergency U.S. government order over cybersecurity concerns.
Maybe the concern was valid after all.
It remains to be seen if this bizarre cyber attack stays as a one-off instance or the start of something far worse.
Until then, a close eye should always be held when tooling with the newest advancements of tech.
Has The Matrix and Terminator taught us nothing?
