🛠 Anthropic brings Mythos 5 to Claude Security
Scanning in Claude Security runs on Claude Mythos 5 — the most powerful model, public beta for Enterprise. The model returns each finding with a CWE category, confidence, severity, and a ready-made fix. There is no direct access to Mythos 5 — only findings and patches go out, each approved by a human. Billing is like regular tokens.
🌍 Anthropic shows how the market will receive cyber capabilities from a frontier model: without direct API, only defensive artifacts — findings and patches — and an expanded Cyber Verification Program. This pattern will likely become the standard.
👤 On Claude Enterprise — already in the admin console, can be enabled right now. Plus the Defender Advantage fund (0xDAF): $35 million in Claude credits for closing vulnerabilities in open-source.
Source 1: https://claude.com/blog/bringing-claude-mythos-5-to-more-defenders
Source 2: https://claude.com/product/claude-security