Security researchers from Calif, a Palo Alto‑based firm, say they breached macOS by crafting a privilege‑escalation exploit with Anthropic’s Claude Mythos Preview. The team used the AI to spot known‑class bugs, then applied human ingenuity to build the payload. The result is a kernel‑memory‑corruption flaw that lets attackers hijack a MacBook’s core functions for remote control, data theft, and permanent access.

Apple responded by meeting Calif’s researchers at Apple Park and labeling the issue the first public macOS kernel exploit on M1 silicon. The company pledged to patch the vulnerability promptly and stressed security as a priority. Meanwhile, Anthropic’s Project Glasswing, joined by AWS, Google, Microsoft, and others, demonstrates that AI can uncover hidden bugs before malicious actors exploit them.

The incident underscores the growing partnership between security firms and AI platforms to preempt attacks. By revealing a kernel‑level flaw that could grant full system control, Calif’s work forces Apple to accelerate its patch cycle and highlights the risk of AI‑assisted exploits. As companies adopt defensive tools, the race to secure silicon will demand collaboration between hardware and software teams.