First Apple M5 memory exploit discovered using Anthropic AI, gives root access on MacOS — Claude Mythos helps security researchers bypass Memory Integrity Enforcement

A new memory exploit, the first for Apple M5 chips, has been discovered using Anthropic AI, granting root access on MacOS by bypassing Memory Integrity Enforcement. The exploit, discovered by security researchers at Calif, allows a standard user to gain administrator access on the machine. The vulnerability was disclosed to Apple in advance and is part of a series of security findings aided by AI tools. The exploit chain successfully evades MIE, a security feature on M5 chips that enforces memory integrity checks at the hardware level. This discovery highlights the ongoing challenges in securing operating systems against evolving threats.