Claude Mythos linked to alleged M5 kernel exploit in 5 days

An Instagram post from @techsinlast24hrs claims a Palo Alto startup leveraged Anthropic’s restricted Claude Mythos preview to produce a macOS kernel memory corruption exploit targeting Apple’s M5 chip [@techsinlast24hrs]. The exploit allegedly granted root access on macOS 26.4.1 from an unprivileged user account, bypassing Apple’s memory integrity protections.

According to the post, the team achieved the breach in five days. Apple has spent nearly five years developing the M5’s security architecture, making the speed of this reported exploit notable—if confirmed. The post does not include technical proof, reproducible steps, or independent verification.

Claude Mythos is not publicly available. Access is limited to select partners under strict use policies, according to Anthropic. The company has not commented on whether the model was involved in generating exploit code or whether the output violates its safety guidelines.

Apple also has not responded to the claim. Kernel-level memory corruption exploits on Apple’s ARM-based silicon are rare, particularly within days of a chip’s release. Prior public exploits on M-series hardware have taken months or years to surface.

The claim remains unverified and rests entirely on a single social media post. No code, logs, or forensic analysis have been shared. Until independent experts review the methodology, the report should be treated as speculative.