Particle.news
Download on the App Store

U.S. Restores Limited Access to Anthropic's Claude Mythos 5

The decision reflects a new U.S. practice of pre-release federal review that could change how advanced AI models are launched and governed

Overview

  • Citing reported security bypasses, Commerce Secretary Howard Lutnick ordered an abrupt restriction of Anthropic’s Claude Mythos 5 and Fable 5 on June 12.
  • After two weeks of technical work with U.S. agencies, Lutnick on June 27 conditionally reauthorized Mythos 5 for more than 100 vetted U.S. government agencies and companies.
  • Fable 5, the consumer-targeted variant, remains blocked and was not included in the conditional restart.
  • Anthropic and federal officials say they fixed vulnerabilities that had allowed safeguards to be bypassed, a process that included addressing issues flagged by Amazon and other security researchers.
  • The move is already changing industry practice by prompting client-by-client rollouts at rivals such as OpenAI and raising questions about reliance on U.S. providers versus open or non-U.S. models.