White House blocks Anthropic's Claude Mythos over export control concerns tied to SK Telecom
Tags AI · Security · Policy
The White House ordered Anthropic to revoke SK Telecom's access to its Claude Mythos cybersecurity model, citing alleged ties to China. Days later, Anthropic took the model offline entirely. The Trump administration has indicated that Anthropic cannot rerelease Mythos or Fable 5 unless it can guarantee the models' guardrails cannot be circumvented — a condition security experts say is technically impossible. The episode highlights the tension between AI export controls and the practical reality that AI models, unlike physical goods, cannot be meaningfully contained once distributed.
Technical significance
The Mythos controversy illustrates a fundamental challenge in AI governance: unlike traditional dual-use technologies, AI models are software that can be copied and distributed globally at near-zero cost. The administration's demand for unbreakable guardrails conflicts with the well-established reality that all AI safety measures can be circumvented, raising questions about whether export control frameworks designed for physical goods can meaningfully regulate AI.