NSA testing finds Mythos vulnerable
Anthropic’s Mythos model reportedly breached the defenses of nearly all NSA systems during red-team trials. The tests were part of a risk assessment ahead of lifting export restrictions.
The Trump administration had banned foreign nationals from accessing Mythos 5 and Fable 5 due to security concerns. The new results suggest current safety protocols are insufficient against a model of this complexity.
Access to Mythos 5 and Fable 5 remains restricted to non-US persons. Anthropic is still in talks with the White House to ease these restrictions, but the NSA test results are likely to strengthen regulators’ hands.
Up next
Nff turns Claude into a firmware development engineer
Claude can now flash microcontrollers and fix them in the field. The nff tool gives the model USB access for compiling, writing, and debugging code in real-time.
Related stories
- Alberta government scans 466 million lines of code with Claude
- China reaches parity with Anthropic in vulnerability research
- Trump administration restricts access to Anthropic models
- Anthropic talks with White House over Fable 5 stall
- Apiiro finds Claude Code secrets in 1,122 GitHub files
- Email leak stopgap in Claude Code checks only shell calls
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.