Even Apple's shiny new M5 fortress isn't immune to a little AI-powered lock-picking. A startup called Calif just proved that artificial intelligence can shrink months of complex hacking into a mere few days. Using a preview of Anthropic’s Claude Mythos model, researchers Bruce Dang and Dion Blazakis managed to bypass Apple’s sophisticated Memory Integrity Enforcement system between late April and early May. This hardware-backed security was meant to be the ultimate shield against memory attacks, yet the AI acted as a super-charged co-pilot to identify weaknesses at lightning speed. While a 55-page report is currently sitting on a desk in Cupertino, there’s no need to panic and sell your Mac just yet. This was a controlled, responsible disclosure, and a patch is almost certainly on the way. The real takeaway? In the race between hackers and defenders, AI just shifted everyone into high gear.