Anthropic's most capable public model arrived June 9 with access controls built to keep China's AI labs out.
Claude Fable 5 is the public version of Mythos, a model Anthropic had held back since April over what it could do. The restrictions target China's AI labs specifically. The backlash was harder to predict: criticism arrived quickly, and not from Beijing, but from Anthropic's own side of the divide.
That internal friction is the sharper signal here. Geographic restrictions on AI models follow chip-export logic: limit what rivals can reach and accept the collateral. When the loudest objections come from your own community rather than the intended target, it suggests the controls are landing somewhere the policy didn't fully account for.
Sitting on a model for two months and shipping a constrained public version is starting to look like standard operating procedure for frontier AI labs.