Particle.news

U.S. Reauthorizes Anthropic’s Top AI Models After Security Pledge

Anthropic agreed to detect security flaws and cooperate with the Commerce Department, allowing export controls to be lifted under conditions that keep government oversight in place.

Overview

  • The Commerce Department forced Anthropic to suspend access to its most powerful models, Mythos 5 and Fable 5, on June 12 after regulators said they found gaps in the models' safety guardrails.
  • After negotiations, Anthropic regained limited access for a small group of U.S. cyber defenders and infrastructure operators before the full restrictions were reviewed.
  • On June 30 the Commerce Department lifted its export controls after Anthropic committed in a public letter from Secretary Howard Lutnick to proactively detect and mitigate security risks and to work with government protocols, with the U.S. reserving the right to reimpose limits.
  • Rivals and vendors adapted quickly: OpenAI released GPT-5.6 in very limited form under government client-by-client validation, and Anthropic said it will work with Amazon, Microsoft and Google on a common framework to rate the severity of AI security flaws.
  • The episode has intensified calls for digital sovereignty in Europe, prompted France to pledge €655 million for AI, and left uncertainty over whether foreign state partners will regain access to Anthropic's models.