U.S. Export Controls Force Anthropic to Pull Fable 5/Mythos 5 AI Models Worldwide

Washington’s snap clamp-down recalls the 1949–1950 birth of Cold-War CoCom rules that abruptly barred western companies from selling strategic tech—then vacuum-tube computers—to the Soviet bloc. Like that earlier embargo, today’s order signals that intangible code can be deemed a munition once its strategic value crosses a perceived threshold. The action fits a decade-long trend: the U.S. has shifted from laissez-faire AI rhetoric (2016) to chip export bans (2022) to now policing model weights themselves, steadily broadening the national-security perimeter. Over a 100-year horizon this moment matters because it sets precedent for states to gatekeep knowledge flows, fragmenting the global compute commons much as nuclear export regimes splintered physics after 1945. Whether the restriction sticks or is relaxed, the Rubicon—government assertion of veto power over AI releases—has been crossed, altering the trajectory of how, where, and by whom frontier intelligence is developed and shared.