Anthropic's Mythos AI Finds Apple macOS Security Flaw

Anthropic's AI, Mythos, is already uncovering security flaws in Apple software. This AI is so effective at finding vulnerabilities that it's not publicly available; instead, a select group of security researchers and large organizations have access. For example, security researchers used Mythos to find a crack in Apple's macOS. They found an exploit that could give an unprivileged local user complete access to a device. This exploit involved two vulnerabilities and several techniques. Mythos helped them discover these bugs and assisted with exploit development. What's interesting is that Mythos can generalize its learning to new problems within a known class of bugs. It's currently unclear if Apple has patched this specific flaw. However, Apple's release notes for macOS Tahoe 26.5 mention a fix for a bug submitted by Calif in collaboration with Claude and Anthropic Research. This shows the significant potential of AI in enhancing cybersecurity efforts.