Within 72 hours of the 9 June release, the US government issued an export-control directive requiring Anthropic to restrict foreign nationals’ access to Claude Fable 5 and Mythos 5 — and, the company said, the practical effect was that it had to disable access globally for all users.
The Anthropic Fable 5 and Mythos 5 suspension
The swift removal of Anthropic’s Claude Fable 5 and Mythos 5 models illustrates how sovereign decisions can instantly reshape market access. According to the source, a US export-control directive aimed at restricting foreign nationals’ access led Anthropic to disable access worldwide so it could comply. The article frames the result bluntly: a leading-edge AI capability that was available to enterprises and governments worldwide was removed from the market by sovereign decision.
The episode is used in the source to establish three linked facts: frontier AI models are being treated as strategic assets; mechanisms of restriction can produce global effects even when targeted at specific populations; and access to these capabilities is contingent, not guaranteed.
U.S. sanctions and the International Criminal Court, 2025
The article pairs the Anthropic case with a 2025 episode in which US sanctions on an International Criminal Court prosecutor reportedly interfered with the tribunal’s access to technology platforms, disrupting its ability to function in practice. The source draws the parallel to show a familiar mechanism: a sovereign leveraging jurisdiction and regulatory power to restrict the operation of systems that in theory operate above national politics. In both instances, the outcome was not negotiated — it was imposed.
TPDI’s "Expanding AI Sovereignty to AI Agency" — framework and critique
The Tech Policy Design Institute (TPDI) released a report this month titled Expanding AI Sovereignty to AI Agency, which the article describes as providing a “well-structured framework” for understanding Australia’s place inside a globally interconnected AI system. The article, however, argues the report “ultimately rests on a set of assumptions that are now visibly out of step with geopolitical reality.”
Central to TPDI’s framing is the notion that countries such as Australia can shape outcomes through influence and interdependence — that is, through leverage rather than direct control. The article notes that policy translation of this idea tends to assume bargaining as the primary tool: conditioning access to domestic strategic assets, such as critical minerals, to secure continued access to foreign-controlled AI capabilities.
Policy choices: where sovereign control matters and why
The article counsels a distinction between optimisation and assurance. When sovereign authority can revoke access quickly, optimisation through diversification and partnerships “presupposes stability” and therefore is an insufficient floor for continuity. For Australia, it recommends first identifying technologies and capabilities for which sovereign control is preferred or essential, then aligning procurement, investment and regulatory settings to build and maintain those capabilities domestically.
That approach, the article says, does not reject interdependence but recognises its limits: “Technology policy is not an abstract exercise in positioning. It is about ensuring that the systems on which government, society and the economy depend can operate reliably under all conditions. This is a question of resilience, not efficiency.”
What this means for policymakers, procurement leaders, and technologists
- Policymakers: Decide which layers of the AI ecosystem are non-negotiable and commit procurement and investment to domestic control for those layers, rather than relying primarily on leverage and bargaining.
- Procurement and enterprise leaders: Treat access to frontier capabilities as contingent; align buying and supply choices to prioritise continuity of capability over marginal advantage.
- Technologists and security teams: Design systems on the assumption that external access can be revoked, and prioritise resilience in architectures and dependencies that matter for core government and societal functions.
The article’s final, pointed judgment is stark: “Agency that depends on bargaining is not sovereignty; it’s managed dependency.” In the face of cases where access can be switched off quickly and globally, the article argues Australia — as a middle power — must decide where independence is essential and build assurance into those critical layers rather than leave continuity to negotiation and hope.
