Moltbook's real risk isn't AI. It's your data

A Palo Alto security firm said Anthropic’s closed Claude Mythos Preview helped uncover and develop an exploit for a macOS kernel memory corruption flaw on Apple M5 devices.

Apple appears to have fixed at least one related issue, though the full technical details have not yet been disclosed.

The case underscores how frontier AI can speed up offensive security research and make vulnerability discovery more scalable before patches are released.