Particle.news
Download on the App Store

U.S. Orders Anthropic to Block Foreign Access to Its Top AI Models

The Commerce Department cited jailbreak vulnerabilities as officials and Anthropic hold daily technical talks to decide whether the models can be restored.

Overview

  • Commerce officials issued an export‑control directive that told Anthropic to obtain a license and prevent any foreign‑person access to Fable 5 and Mythos 5, a move communicated around June 12–13 that forced immediate action.
  • Anthropic complied by disabling both models for all users worldwide to ensure full compliance rather than try to distinguish U.S. from non‑U.S. users in real time.
  • Reporting says Amazon researchers demonstrated prompt‑based 'jailbreaks' that could coax Fable 5 to reveal information useful for cyberattacks, a finding U.S. officials and agencies including the NSA considered when raising national security concerns.
  • Legal and policy experts say the Commerce letter may be shaky because digital API access does not clearly fit traditional export rules under the Export Administration Regulations, and bipartisan House members have demanded formal explanations.
  • The dispute has prompted industry pushback and diplomatic pressure from allies seeking 'trusted partner' access, and Anthropic on June 18 proposed closer collaboration with the government as daily negotiations continue to work toward a fix.