A research team met at Apple Park this week to hand‑deliver a vulnerability report describing the first public macOS kernel memory‑corruption exploit that survives Memory Integrity Enforcement on M5 silicon. The exploit reaches root from an unprivileged user on macOS 26.4.1, using only standard system calls and two newly discovered bugs.

The team used Mythos Preview, an AI‑driven platform that generalizes across known bug classes, to locate the flaws in days. Bruce Dang found the first bug on April 25, Dion Blazakis joined two days later, and by May 1 the full privilege‑escalation chain worked. Researchers point out Apple spent five years and billions on MIE, yet the exploit needed only a one‑year domain budget.

A 55‑page technical report will be released once Apple patches the two vulnerabilities. By breaking MIE on bare‑metal M5 hardware, the researchers demonstrate that AI‑assisted exploit development can still outpace even the most expensive hardware mitigations. The finding forces developers to treat memory‑tagging as a delay, not a guarantee, and underscores the need for layered defenses.

Presenting the exploit in person rather than through the crowded Pwn2Own submission pipeline gave the team a brief moment of notoriety on Twitter, a rarity among seasoned hackers who avoid direct contact. Their visit also highlighted Apple’s $5 billion campus investment, contrasting with the modest cost of their own operation. The episode illustrates how small, AI‑enhanced groups can challenge the resources of tech giants.