Security research firm Calif used Anthropic’s unreleased AI model, Mythos, to discover a serious vulnerability in Apple's macOS.
The exploit links two software bugs to bypass Apple’s Memory Integrity Enforcement. Apple spent over five years engineering this specific security feature to prevent such attacks.
This privilege escalation technique corrupts Mac memory to gain access to protected system areas. While the AI model identified bugs and assisted development, Calif emphasized that human expertise was essential to build the final exploit.
Calif delivered a 55-page report detailing the findings to Apple. Apple is currently reviewing the research to develop a security patch.