Security research firm Calif has disclosed a significant vulnerability in Apple's M5 chip, achieving full privilege escalation from local user to root shell on macOS devices in just five days. The attack, detailed in a 55-page report, exploited two zero-day kernel vulnerabilities and a side-channel tag collision technique to bypass Apple's MIE hardware memory safety mechanism. This marks the first public exploitation of macOS kernel memory corruption targeting the M5 chip's MIE. The research was accelerated by the AI tool Claude Mythos Preview, which played a crucial role in identifying vulnerabilities. Following the disclosure, Apple has patched the vulnerabilities in macOS 26.5. The breakthrough highlights the growing role of AI in cybersecurity, with both defenders and attackers leveraging AI to accelerate vulnerability discovery and exploitation.