AI helped researchers bypass Apple M5 defenses

Security startup Calif says researchers used Anthropic’s Claude Mythos Preview to help develop a macOS kernel exploit that bypassed Apple’s Memory Integrity Enforcement on M5 silicon in under a week, targeting macOS 26.4.1 and achieving root access via local privilege escalation. The claim highlights how AI can accelerate vulnerability discovery, raising short-term reputational and operational risk for Apple (AAPL.OQ) as customers and enterprises scrutinize device security and await a patch. While Apple’s security model remains robust and the issue requires initial code execution to be exploited, the story could increase pressure for faster patching, disclosure practices and potential regulatory interest — factors that may modestly weigh on investor sentiment until fixes are deployed.