US government forces Anthropic to disable Claude Fable 5 and Mythos 5 for all customers worldwide

The selective nature of the shutdown is the detail that matters most. If jailbreak vulnerabilities in GPT-5.5 are materially similar and face no equivalent enforcement, this action functions less like a technical safety intervention and more like a regulatory chokepoint applied unevenly, which raises questions about the criteria actually driving the decision.

Modelwire has no prior coverage to anchor this to directly. This story belongs to an emerging category we have not yet tracked systematically: government-initiated model suspensions as a distinct enforcement mechanism, separate from the voluntary pause agreements and safety commitments that labs have publicized over the past two years. The closest relevant context is the broader pattern of frontier labs using safety positioning as competitive differentiation, a dynamic this action complicates considerably for Anthropic specifically.

Watch whether the US government issues any formal public criteria for what constitutes an actionable jailbreak vulnerability within the next 60 days. If no criteria are published, the enforcement action will be very difficult to distinguish from discretionary regulatory pressure, and other labs will have no reliable basis for assessing their own exposure.