Researchers Use Anthropic Mythos to Expose Apple macOS Security Bugs

▤ Summary

Researchers used Anthropic’s Claude Mythos Preview to help identify security flaws in Apple macOS, including a reported kernel memory corruption exploit on Apple M5. Calif said Mythos assisted with bug discovery and exploit development but that skilled human researchers guided and completed the process. The reported attack chain involved two separate vulnerabilities that could be chained to corrupt memory, and full technical details remain private until Apple addresses the issues. Apple is reviewing the report and the patch status remains unclear, and the case raises concerns about how AI tools can accelerate both defensive research and potential offensive use.

▦ Linked Companies