Three days after Anthropic confirmed it was pulling Fable 5 and Mythos 5 from global access under a Bureau of Industry and Security directive, allied governments are beginning to respond, and the framing has shifted from national security enforcement to something broader.
Canadian Prime Minister Mark Carney reportedly addressed the directive around June 14, pointing to concentration risks in US AI supply chains, according to the Associated Press. The specific wording of his remarks could not be independently confirmed before publication. But the general shape of the concern, that a single government decision can sever access to frontier AI infrastructure for the entire world, is now on the record at the head-of-government level.
That’s the accountability question no one had formally asked before June 12.
The directive itself is established. CNBC reported that Anthropic disabled both models globally rather than selectively blocking foreign nationals, because the deemed export rule, which treats sharing controlled technology with a foreign national anywhere as an export, made selective enforcement impractical inside its own workforce. A company with non-US employees had no clean technical path to compliance other than a full shutdown. The models had launched June 9. The directive came three days later.
Governmental Responses to US AI Export Control Authority
The Commerce Department’s action was reportedly triggered by a third-party jailbreak demonstration of Fable 5’s cybersecurity features, according to Axios reporting, the technical details haven’t been officially disclosed. Anthropic reportedly had approximately 90 minutes to pull the models, according to media reporting, though that timeline hasn’t been confirmed by official government sources.
What the Carney statement, even in its unconfirmed form, signals is that Washington’s AI export control authority is now being evaluated by trading partners as a supply chain risk, not just a national security instrument. Canada’s exposure is real: Canadian enterprises and government agencies that integrated Fable 5 or Mythos 5 APIs lost access with the same 90-minute window as any other foreign customer. There was no diplomatic carve-out, no allied-nation exception, no advance notice to Ottawa.
This isn’t the first time allied governments have raised AI concentration concerns. Japan’s Digital Minister warned of “AI colony” risk earlier this month as the Diet debated consent-free data processing rules. The EU’s Coordinated AI Development Agreement includes sovereignty tier provisions specifically designed to reduce dependency on non-European foundation models. Those frameworks were theoretical. The June 12 directive made them operational questions.
The catch is that the regulatory response options are limited. Building a domestic frontier model to replace Fable 5-class capabilities takes years and billions. Diversifying to non-US providers is possible but narrows the capability set. Negotiating allied-nation exemptions from US export control authority, the path Canada would likely prefer, has no established mechanism in current BIS rules for AI models.
Who This Affects
The real question is whether June 12 becomes a one-time enforcement action or the opening case in a sustained pattern of BIS asserting jurisdiction over AI model access. Multinational compliance teams with foreign national employees using US frontier APIs should treat this as a scenario-planning trigger now, not a policy question for later. The deemed export rule existed before Fable 5. It will exist after this case resolves.
If allied governments formalize their concentration risk assessments into procurement policy, which Canada, the EU, and Japan all have the institutional capacity to do, US AI providers could face a fragmented access landscape for their most capable models. That’s the downstream consequence that doesn’t show up in the current coverage of this story.