The U.S. government ordered Anthropic to block foreign access to its new Fable 5 and Mythos 5 AI models after officials said a sensitive jailbreak method posed national‑security risks. Anthropic pushed back, arguing the capability already exists in other public models and that the shutdown was an overreach driven by export‑control rules. The dispute highlights growing tension between AI developers and federal regulators over how to handle advanced model risks.