On June 12, 2025, the US Department of Commerce’s Bureau of Industry and Security (BIS) issued an export control directive requiring Anthropic to suspend all foreign national access to its AI models Mythos 5 and Fable 5, both domestically and abroad. This action represents a groundbreaking legal precedent applying export controls to continuously accessible AI models delivered via API.
What Happened
The Bureau of Industry and Security issued a directive to Anthropic on June 12, 2025, mandating that it block any foreign national—regardless of location—from accessing its AI models Mythos 5 and Fable 5. Unable to reliably screen users by nationality, Anthropic disabled both models entirely. The measure was taken on national security grounds following reports of “jailbreaking” vulnerabilities and concerns about a China-linked actor potentially accessing the models.
Key Facts
- Jurisdiction: United States, under the Department of Commerce
- Agency: Bureau of Industry and Security (BIS)
- Directives issued June 12, 2025
- Models affected: Anthropic’s Mythos 5 and Fable 5
- Restriction: Suspension of foreign national access to these models regardless of location (inside or outside the US)
- Enforcement basis: National security concerns under the Export Administration Regulations (EAR)
- Anthropic disabled the models entirely due to inability to filter user nationality
- The directive letter has not been publicly released
Why It Matters
This enforcement action signals a shift in how US export control law applies to AI technology, extending controls from tangible goods and discrete software transfers to continuously delivered frontier AI services accessed globally via APIs. This poses challenges for compliance and enforcement since traditional export controls focus on identifiable transfers, while cloud-hosted AI models are perpetually accessible. The case also raises questions about which framework—incremental-risk or capability-based—will govern future AI export controls, impacting the US AI sector and international competition.
Background
US export controls have historically covered not only physical goods but also intangible software and technical data under the Export Administration Regulations. Past policy debates on encryption software from the 1990s affirmed that code and software can be subject to such controls. However, this is the first notable application to AI models operating as continuously accessible online services, rather than standalone software copies or physical transfers.
Analysis
Policy analysts note this enforcement action establishes a precedent for regulating AI models by controlling access rather than distribution of discrete files. The government faces a choice between an incremental-risk approach—limiting controls to genuinely novel AI capabilities—and a broader capability-based approach that targets any presence of sensitive AI functionalities. Experts caution that while the capability-based approach is easier to administer, it may impose burdens on US firms without effectively denying adversaries access. Conversely, the incremental approach demands deep real-time insight into global AI capabilities, posing institutional challenges.
Who Is Affected
- Anthropic as the AI developer directly impacted
- Foreign nationals worldwide barred from accessing the specified AI models
- US AI companies facing potential export control risks for frontier models
- US government regulators and policymakers shaping AI export rules
What Remains Unclear
- The specific content and scope of the Commerce Department’s directive remain undisclosed publicly
- Details surrounding the alleged jailbreak incident prompting the action are disputed
- Whether the government is applying an incremental-risk or capability-based enforcement test
- How broadly this precedent will be applied to other AI models and companies
- The full role and influence of national security concerns versus export control technicalities
What Comes Next
This information was not confirmed in the reviewed sources.
Sources
This article is based on reporting and publicly available information from the following source:
Read more AI Regulation stories on Goka World News.