The NSA is reportedly using Anthropic's restricted cybersecurity AI model Mythos Preview for vulnerability scanning, despite the Pentagon previously labeling Anthropic a 'supply-chain risk' due to the company's refusal to allow unrestricted military access. This development comes amid a thawing relationship between Anthropic and the Trump administration, with CEO Dario Amodei recently meeting with White House officials. The model was withheld from public release due to its potential offensive capabilities and is only available to about 40 selected organizations.
Background
Anthropic recently developed Mythos, a frontier AI model designed for cybersecurity tasks but deemed too dangerous for public release due to offensive capabilities. The Pentagon had previously clashed with Anthropic over access to its AI models for military applications.
- Source
- TechCrunch
- Published
- Apr 21, 2026 at 12:06 AM
- Score
- 8.0 / 10