Researchers Build First Public Apple M5 macOS Kernel Exploit with Mythos Preview

Security researchers have unveiled the first publicly known macOS kernel memory corruption exploit targeting Apple’s latest M5 silicon, marking a significant moment for both offensive security and Apple’s next-generation defenses.

The exploit, developed in collaboration with Mythos Preview, reportedly bypasses Apple’s advanced Memory Integrity Enforcement (MIE), a hardware-backed mitigation designed to stop this class of attacks precisely.

The research team disclosed the findings directly to Apple during a visit to Apple Park on May 14, 2026, opting for in-person reporting instead of traditional submission channels.

Full technical details remain under embargo until patches are released. However, early insights reveal a sophisticated attack chain built in just five days.

Apple M5 macOS Kernel Exploit

Apple introduced MIE alongside its M5 and A19 chips as a flagship security feature. Built on ARM’s Memory Tagging Extension (MTE), MIE enforces strict memory safety rules at the hardware level, aiming to eliminate entire classes of memory corruption vulnerabilities.

Apple previously claimed the technology could disrupt all known exploit chains targeting modern iOS devices. Despite this, researchers demonstrated that MIE is not foolproof.

The exploit targets macOS 26.4.1 (build 25E253) and achieves a full kernel compromise using a data-only attack chain.

Starting from an unprivileged local user, the exploit relies solely on legitimate system calls and escalates privileges to root. Notably, it avoids traditional code-execution techniques and instead manipulates kernel memory structures to achieve control.

A key element in the breakthrough was Mythos Preview, an AI-assisted vulnerability research system. The platform helped identify underlying bugs and supported the development of exploits.

• Mythos rapidly detected vulnerabilities belonging to known bug classes.
• Human researchers refined the approach to bypass MIE protections.
• The final exploit chain combined two vulnerabilities with multiple advanced techniques.

This hybrid workflow highlights a growing trend in cybersecurity: AI systems accelerating vulnerability discovery. At the same time, human expertise remains critical for bypassing complex mitigations.

For example, Mythos can generalize across bug classes, meaning that once it learns to identify one type of flaw, it can scale that knowledge to similar targets. However, bypassing hardware-enforced protections like MIE still requires creative, manual exploitation strategies.

Calif researchers describe the work as a preview of what’s ahead. As AI tools continue to evolve, the speed and scale of vulnerability discovery are expected to increase dramatically. This raises concerns that even advanced mitigations, such as MIE, may face growing pressure from AI-assisted exploit development.

Importantly, the exploit demonstrates that while Apple has significantly raised the bar, determined attackers can still find viable paths, especially when combining automation with deep technical expertise.

To date, this is believed to be the first public exploit capable of bypassing MIE on macOS running M5 hardware.

Apple is expected to patch the vulnerabilities in an upcoming security update. The research team plans to release a detailed 55-page technical report once fixes are available, which will likely provide deeper insight into both the exploit chain and MIE bypass techniques.

The findings reinforce a critical reality in modern cybersecurity: no mitigation is absolute. As defenders deploy increasingly sophisticated protections, attackers are evolving just as quickly, now with AI as a force multiplier.

Follow us on Google News, LinkedIn, and X to Get Instant Updates and Set GBH as a Preferred Source in Google.