First public macOS kernel memory corruption exploit on Apple M5

· security systems · Source ↗

TLDR

  • Calif security team, aided by Mythos Preview AI, built a working MIE-bypassing macOS kernel LPE exploit on M5 silicon in five days.

Key Takeaways

  • Exploit is a data-only kernel local privilege escalation chain on macOS 26.4.1, starting from an unprivileged user and ending with a root shell.
  • MIE (Memory Integrity Enforcement), Apple’s hardware-assisted memory safety system built on ARM MTE, is the marquee security feature of M5 and A19 chips.
  • Two vulnerabilities discovered by Bruce Dang on April 25; working exploit delivered by May 1, with Mythos Preview assisting bug discovery and Dion Blazakis and Josh Maine building tooling.
  • Mythos Preview generalizes across known bug classes quickly but still required human expertise to navigate MIE bypass, which is a novel mitigation class.
  • Full 55-page technical report will be released after Apple ships a patch; coordinated disclosure was delivered in person at Apple Park.

Hacker News Comment Review

  • Commenters broadly flagged the post as a PR piece with no verifiable technical details yet; skepticism is high until the 55-page report drops.
  • There is debate over whether Mythos deserves credit or whether the result reflects elite human researchers using any capable model as a force multiplier; the Nicholas Carlini comparison was raised.
  • The exploit class matters: commenters clarified this is a local privilege escalation, not a zero-click RCE, putting its bug bounty value around $100K rather than $1.5M unless repackaged against a locked-mode beta.

Notable Comments

  • @yieldcrv: Pegs current bounty value at ~$100K but notes a path to $1.5M if demonstrated as unauthorized access against a beta build with Lockdown Mode.

Original | Discuss on HN