US greenlights Claude Mythos 5 release

The resolution of the Mythos block shows that the US government is actively gatekeeping frontier AI releases, creating a two-tiered access system for national security-sensitive models.

• –Access is restricted to over 100 vetted US corporations and government agencies under Anthropic's Project Glasswing cybersecurity consortium.
• –Claude Mythos 5 is designed for highly advanced tasks in cybersecurity, software engineering, and scientific research, where its agentic capabilities exceed human benchmarks.
• –Anthropic's consumer-facing Claude Fable 5 remains under discussion for a wider release, utilizing safety classifiers to route high-risk queries to older models.
• –This intervention sets a clear precedent for how the US government plans to exercise export controls and national security oversight on frontier LLMs.