Particle.news

U.S. Bars Foreign Access to Anthropic’s Top AI Models

The export-control order treats frontier AI as a strategic, state-controlled asset and signals a shift in how advanced models will be governed

Overview

  • U.S. officials have issued an export-control directive that suspends foreign access to Anthropic’s Fable 5 and Mythos 5, and the restriction currently remains in effect.
  • The government cited national security concerns after reports that researchers could use 'jailbreak' techniques to bypass Fable 5’s built-in safety limits and extract sensitive outputs.
  • Anthropic says no public demonstration has proved a bypass and company leaders will meet Commerce Department officials to contest the directive.
  • More than 50 cybersecurity leaders, including executives tied to major tech firms, sent a formal letter asking the government to reverse the ban because it could reduce the industry’s ability to find and fix software vulnerabilities.
  • France announced an extra €655 million for national AI programs in direct response, highlighting immediate geopolitical pressure for countries to build independent AI capacity and new rules for access to advanced models.