LLM

In early May 2024 a joint effort by the security firm Anthropic and the research team at Calif demonstrated that a sophisticated local‑privilege‑escalation (LPE) exploit could be built against macOS 26.4.1 running on Apple’s newest M5 silicon. The key differentiator was the use of Anthropic’s large‑language model (LLM) Mythos Preview, which dramatically compressed the discovery and development cycle that traditionally takes months into a matter of days.