Anthropic's Mythos Reveals the Geopolitics of AI Withholding

Anthropic's framing around Mythos positions the access restriction as a defensive posture — the model is capable of finding vulnerabilities in software and computer systems at a level that makes broad release irresponsible. That framing is credible. It is also structurally identical to an export control. Whether Anthropic intended the geopolitical consequence is beside the point — the consequence is already in motion, and allied governments that did not make the list have no mechanism to appeal it.