Particle.news

U.S. Lifts Export Controls on Anthropic’s Most Powerful AI Models

The Commerce Department’s move allows Anthropic to begin restoring access and reflects a new U.S. practice of vetting frontier AI for national‑security risks.

Overview

  • The U.S. imposed export controls that forced Anthropic to cut access to Mythos 5 and Fable 5 on June 12, citing security flaws that could enable cyberattacks and other misuse.
  • After limited talks with Washington, Mythos 5 was partly reinstated for a narrow set of U.S. cyberdefense and infrastructure operators before the full controls were lifted.
  • The Department of Commerce lifted the export controls on June 30–July 1 and Anthropic said it will begin restoring access to Mythos 5 and Fable 5, though details on foreign partner access remain unclear.
  • CIA Director John Ratcliffe publicly likened top AI models to “digital nuclear weapons” while AWS announced a classified cloud offering and $1 billion in credits for intelligence contractors, showing deeper public‑private coordination.
  • The episode sets a new precedent where the U.S. can force commercial model withdrawal then conditionally restore service, raising legal questions about export rules, allied access, and who will be allowed to use future frontier models.