What happened
The National Security Agency is using Anthropic's most powerful Mythos Preview model despite DoD officials labeling the company a 'supply chain risk' and cutting off contracts in February. The NSA is among unnamed agencies with access to the restricted model.
Why it matters
This reveals internal government conflicts over AI vendor risk assessment and demonstrates how cybersecurity capabilities may override supply chain concerns, potentially setting precedent for other restricted AI tools.
Action needed
Review current AI vendor risk assessments and prepare for potential policy shifts regarding restricted AI model access for critical security functions.