NSA Uses Anthropic’s Mythos Model Despite Dispute with the Pentagon

Colleagues, a note from the AI sector: Axios reports the NSA has obtained access to Anthropic's Mythos Preview.

- Anthropic restricted Mythos access, deeming the model potentially dangerous and providing it to roughly 40 organizations.
- The Pentagon previously described Anthropic as a 'supply‑chain risk' over access limits to model capabilities.
- The NSA reportedly uses Mythos to scan environments for vulnerabilities; the UK AI Security Institute has confirmed similar access.

Why this matters: broader use of powerful AI in defence raises questions about access controls and exploitation risks.

How do you assess the balance between defensive access and privacy/security risks?

#AI #cybersecurity #Anthropic #governance