Particle.news

U.S. Blocks Foreign Access to Anthropic’s Most Advanced AI Models

U.S. officials cite reported jailbreaks that could defeat safety controls as a national security justification for the export order.

Overview

  • The Commerce Department issued an export‑control directive in mid‑June that suspends access by any foreign national to Anthropic’s Fable 5 and Mythos 5, and Anthropic has temporarily disabled those models to comply.
  • U.S. authorities say researchers demonstrated so‑called jailbreak techniques that can bypass model safeguards, a claim Anthropic disputes and says has not been proven to its satisfaction.
  • Reports that Amazon researchers flagged vulnerabilities helped prompt the government action, and Anthropic executives are scheduled to meet Commerce Department officials to discuss compliance and next steps.
  • More than 50 cybersecurity leaders and tech executives warned the restriction could weaken defensive security work because those models are used to find and fix software vulnerabilities.
  • European governments reacted by accelerating AI policy and funding moves, including France’s announcement of an extra €655 million for domestic AI capacity, highlighting a shift toward national strategies for frontier models.