Researchers bypass Apple's MIE with first public M5 kernel exploit in five days

Calif, working with an AI system called Mythos Preview, built the first public macOS kernel memory corruption exploit that survives Memory Integrity Enforcement (MIE), Apple’s hardware-assisted memory safety system based on ARM’s Memory Tagging Extension. MIE shipped on the M5 and A19 as the marquee defense against the vulnerability class behind most sophisticated iOS and macOS compromises, and Apple has claimed it disrupts every known public exploit chain, including the leaked Coruna and Darksword kits.

The chain is a data-only local privilege escalation against macOS 26.4.1 on bare-metal M5 hardware with kernel MIE enabled. It starts from an unprivileged user, relies only on normal system calls, and lands a root shell using two vulnerabilities. Bug discovery began April 25 and a working exploit existed by May 1. Mythos Preview generalized known bug classes to find the flaws quickly, while human experts handled the MIE bypass. The team disclosed in person at Apple Park and is withholding the 55-page technical writeup until Apple patches.

The significance is less about a single bug and more about the pairing: a five-year, multi-billion-dollar hardware mitigation defeated in a week by a small team augmented by AI. The authors frame it as an early signal that AI-assisted vulnerability research will increasingly surface bugs powerful enough to punch through even state-of-the-art exploit mitigations.