Anthropic’s Claude Oceanus-v1-p Opens to Red Team Testing, but Distribution is Compromised

**Executive summary:** Anthropic's restricted Claude Oceanus-v1-p model was exposed through unauthorized API proxies and reportedly resold by an unidentified actor via a Chinese-based proxy service; the model, derived from the Claude Mythos line, is assessed to possess strong offensive capabilities including identifying and exploiting zero-day vulnerabilities. Anthropic paused broader red-team access while investigating, noting the risk to newly enrolled Project Glasswing partners across critical infrastructure sectors.