Anthropic's Mythos AI cracks Apple M5's billion-dollar security in 5 days
Researchers used Mythos Preview to find the first public macOS kernel exploit on M5 silicon.
Deep Dive
Researchers from Calif. used Mythos Preview to find the first public macOS kernel memory corruption exploit on Apple's M5 chip. Apple spent five years and an estimated several billion dollars building Memory Integrity Enforcement (MIE), a hardware-assisted memory safety system around ARM's MTE. The researchers built a working exploit in five days. The full 55-page technical report drops after Apple patches the vulnerability.
Key Points
- Researchers used Anthropic's Mythos Preview AI tool to find the first public macOS kernel exploit on M5 hardware.
- The exploit bypassed Apple's Memory Integrity Enforcement (MIE), a multi-billion dollar, 5-year effort based on ARM's MTE.
- The full 55-page technical report will be released after Apple patches the vulnerability; researchers delivered findings in person to Apple Park.
Why It Matters
AI-powered vulnerability research just cut years of security development down to days — rewriting expectations for hardware defenses.