Anthropic’s Fable shutdown and US “trusted access” plans ignite a new AI power struggle—who gets the keys?

Anthropic’s Fable has been shut down, a move framed by observers as a major inflection point for open-source AI and model availability. On the policy side, reporting indicates the US Department of Commerce restricted access to Anthropic’s latest models, raising immediate questions about how quickly capabilities will be re-released, under what conditions, and to which partners. At the same time, Foreign Policy highlights that Anthropic is again in the US government’s crosshairs, suggesting the company is navigating renewed scrutiny over safety, compliance, and strategic risk. Separately, Reuters reports that G7 leaders are discussing a framework for “trusted partners” to gain access to cutting-edge US AI models, implying a coordinated approach to distribution rather than purely bilateral deals. Geopolitically, the cluster points to AI governance hardening into a managed-access contest between the United States and China, with rules, data, and critical inputs like rare earths becoming strategic levers. Brazil’s top foreign policy adviser, Celso Amorim, argues that Latin America and Europe are being sidelined as the competition narrows to two powers, which would reshape bargaining power for countries that could otherwise influence standards and supply chains. The “trusted partners” concept also signals that the West may formalize an access-control regime—effectively turning frontier models into quasi-strategic assets—while limiting diffusion to non-aligned or higher-risk jurisdictions. In this environment, Anthropic becomes a focal point: a private AI developer whose deployment decisions can be used to test the boundaries of national security oversight and allied coordination. Market and economic implications are likely to concentrate in AI infrastructure, cloud compute, and the supply chain for critical materials. If access restrictions tighten, demand for compliant enterprise deployments and government-approved model endpoints should rise, benefiting vendors that can integrate “restricted” models into regulated workflows and compliance tooling. The rare-earth angle matters for downstream hardware—GPUs, networking gear, and data-center components—because any tightening of rare-earth sourcing or export controls can amplify cost pressures across semiconductors and electronics. In currency and rates terms, the most direct effects would be through risk sentiment and capex expectations for AI data centers and defense-adjacent tech, with potential near-term volatility in AI-adjacent equities and cloud infrastructure names as policy uncertainty changes the expected revenue path. What to watch next is whether the Commerce restrictions evolve into a broader licensing regime, and whether Anthropic’s model access is reopened under a clearer “trusted partner” rubric. The G7 discussions are a key near-term indicator: if leaders converge on eligibility criteria, audit requirements, and enforcement mechanisms, it would accelerate a two-tier market for frontier AI. Another trigger is whether additional US government actions target other frontier model providers, which would confirm a systemic shift rather than a company-specific dispute. Finally, the rare-earth and data-governance narrative suggests escalation could occur through supply-chain measures or standards battles, so monitoring export-control updates, rare-earth policy moves, and any new AI safety/compliance directives will be essential over the coming weeks.

Geopolitical Implications

• 01

  Frontier AI is being treated as a strategic asset, with access control becoming a tool of alliance management and deterrence.

• 02

  A two-tier global AI market may emerge: government-approved “trusted” deployments versus restricted or delayed capabilities for others.

• 03

  The US-China AI race is expanding beyond models into governance, data flows, and critical-material dependencies, reducing policy room for middle powers.

• 04

  Private AI firms like Anthropic may face recurring national-security oversight that can reshape innovation incentives and deployment geography.

Key Signals

• —Details of the Commerce restriction scope: licensing criteria, audit requirements, and enforcement timelines.
• —Any formal G7 communiqué or technical working-group outputs defining “trusted partners.”
• —Whether additional frontier model providers are subjected to similar access controls.
• —Updates to rare-earth sourcing/export policies and any new data-governance standards tied to AI deployment.