Could AI Help Hackers Find Weak Spots In Your Business?

We've all heard about artificial intelligence doing amazing things, but what about its darker side? A recent experiment by an Aussie security researcher, Kasra, sheds some light on this.

Kasra deliberately built a simple, vulnerable app – like a book review site – and then put various AI models, known as 'Large Language Models' (LLMs), to the test. These are the same types of AI that power tools like ChatGPT. He wanted to see if these AIs could find security weaknesses in the app that a human hacker might exploit. Think of it like setting up a challenge for the AI to find the weak fence post in a digital garden.

The results were mixed, but definitely worth paying attention to. While the AIs didn't immediately become master hackers, they showed a surprisingly good ability to identify potential flaws and even suggest ways to break into the system. It wasn't always straightforward; sometimes the AI needed a bit of nudging or specific instructions. But the fact that they could do it at all, without being explicitly programmed for hacking, is a significant development.

This experiment cost Kasra about $1,500 – mainly for the computing power needed to run these AI tests. This relatively low cost and the AIs' emerging capabilities suggest a future where even less sophisticated cybercriminals could use AI to find vulnerabilities. For everyday Aussie small businesses, this means staying even more vigilant about their online security measures.

It’s a reminder that as AI gets smarter, so do the potential threats. Keeping your business's digital doors locked tight and updated is more crucial than ever. This isn't about panic, but about staying informed and prepared in a rapidly changing digital world.