AI-Assisted Exploit Breaks Apple M5’s Memory Integrity on macOS 26

Researchers used Anthropic's Mythos Preview AI to develop the first public kernel exploit bypassing Apple's M5 Memory Integrity Enforcement in just five days.

CSBadmin
2 Min Read

Exploit Breaches Core Protection

Researchers from Calif, Bruce Dang, Dion Blazakis, and Josh Maine, have developed the first public kernel local privilege escalation (LPE) exploit targeting Apple’s M5 silicon. The attack successfully bypassed Apple’s Memory Integrity Enforcement (MIE), a hardware-level memory protection system built on ARM’s Memory Tagging Extension (MTE) architecture. Running on macOS 26.4.1 (25E253) on bare-metal M5 hardware, the exploit chain starts from an unprivileged local user account, uses only standard system calls, and delivers a full root shell.

The team discovered two underlying bugs on April 25 and had a working exploit by May 1, just five days later. Instead of using standard bug bounty channels, they walked a 55-page printed report directly into Apple Park in Cupertino to avoid crowded submission queues seen during events like Pwn2Own. Full technical details will be published only after Apple ships a patch.

AI Collaboration and Impact

The breakthrough was enabled in part by Anthropic’s Mythos Preview, a powerful AI model that helped identify the two vulnerabilities and assisted throughout exploit development. Calif describes the model as capable of generalizing attack patterns across entire vulnerability classes once it learns a problem type. While the bugs were discovered quickly because they fall within known classes, autonomously bypassing MIE still required significant human expertise, highlighting the power of human-AI pairing.

Memory Integrity Enforcement was introduced as the marquee security feature of the M5 and A19 chips, with Apple spending five years and billions of dollars to engineer it specifically to disrupt kernel memory corruption exploits. Apple’s own research claimed MIE disrupts every known public exploit chain against modern iOS, including leaked Coruna and Darksword exploit kits. The five-day development timeline against a protection that took Apple five years to build is now cited as a significant benchmark for AI-assisted offensive security research, particularly as memory corruption remains the most prevalent vulnerability class across all modern platforms.

Source: Cyber Security News

CSBadmin

The latest in cybersecurity news and updates.

Share This Article
Follow:
The latest in cybersecurity news and updates.