Overview
- Commerce Minister Howard Lutnick confirmed on June 27 that the government found “safeguards” sufficient to let a limited set of trusted partners use Claude Mythos 5 under tight controls.
- Reports say roughly 100 U.S. agencies and companies were approved but the government has not published the list and media accounts suggest some intelligence agencies may have access.
- Anthropic says it will restore Mythos 5 access for organizations that operate and protect critical infrastructure and that it is working with officials to expand access and reinstate Fable 5.
- The move follows Anthropic’s April disclosure that Mythos can identify software vulnerabilities with high precision, a controlled rollout called Project Glasswing, and an export ban tied to a ‘supply‑chain risk’ classification.
- Technology firms have filed court statements calling the controls an overreach, a legal fight that could shape how U.S. export rules govern dual‑use AI and how companies share advanced defensive tools globally.