AI Can Find Computer Glitches No Human Has Seen Yet

Imagine software that's designed to find hidden faults in other software. That's exactly what a new artificial intelligence (AI) tool, named Claude Mythos, has been doing. Created by a company called Anthropic, this AI recently scanned hundreds of open-source projects — which are basically the building blocks for many apps and websites we use daily — and found over 23,000 potential problems.

What's particularly interesting is that many of these problems, or 'bugs', were completely unknown to the human experts who maintain these projects. Think of it like a super-smart digital detective finding tiny cracks in a building that even the best inspectors missed. While not all findings were critical, a significant number of them were acknowledged by the software developers and have even led to real fixes, making the software we rely on more secure.

This isn't about AI taking over human jobs completely, but rather about AI working alongside people to do tasks more efficiently. In this case, Claude Mythos acted like an expert assistant, sifting through massive amounts of code far quicker than any human team could. It identified the areas that needed attention, allowing human security experts to focus their efforts on the most serious issues and confirm the AI's findings.

For everyday Australians, this means the apps, websites, and digital services we use could become more reliable and safer from cyber threats. For small businesses, it highlights how AI can proactively identify vulnerabilities in the digital tools they depend on, potentially saving them from costly security breaches down the track. It's a clear example of AI being put to practical, beneficial use, making our digital world a bit more robust.