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

Washington has not slapped software itself with export controls this aggressively since 1997, when strong encryption was briefly treated as a munition under the Wassenaar regime. Today’s action signals a pivot from policing chips (e.g., the 2022–24 Nvidia/Huawei GPU restrictions) to policing the models that run on them—an echo of 1946’s McMahon Act, which shifted U.S. focus from uranium to the intellectual blueprints of the atomic bomb. The episode also exposes a perennial pattern: private industrial giants (here Amazon) using privileged access to steer security policy, much like AT&T’s role in 1934 radio regulation. On a century horizon, this could mark the moment when sovereigns began treating frontier AI weights as strategic materiel, foreshadowing a splintered global market where capabilities flow only within trusted blocs—potentially slowing the open-source diffusion that has characterized computing since the 1980s, but also birthing a new layer of techno-mercantilism that future historians may compare to early nuclear export regimes.