Calif Builds M5 Mac Root Exploit in Under 1 Week With AI Help

2 articles · Updated · Fox News · Jun 1

Calif said its researchers used Anthropic's Claude Mythos Preview to build a working macOS 26.4.1 kernel exploit on Apple M5 hardware in less than a week, ending with root access.
The startup said the exploit bypassed Apple's Memory Integrity Enforcement, a hardware-backed defense meant to make memory-corruption attacks far harder on A19 and M5 chips.
Calif described the attack as a local privilege-escalation chain, meaning an attacker would still need code execution on the Mac first—such as through a malicious download or compromised installer.
Mozilla has reported similar AI-assisted security gains, saying an early Mythos Preview helped identify 271 vulnerabilities fixed in Firefox 150.
Calif said it disclosed the findings to Apple and plans to publish technical details after a patch, underscoring concerns that AI could compress the time defenders have to fix serious flaws.

With AI defeating Apple's newest chip security in a week, can defenders win an arms race against their own tools?

If AI can bypass billion-dollar hardware security so quickly, is any digital defense system truly future-proof?