In June 2025, the US Commerce Department issued an unprecedented export control directive targeting Anthropic's Mythos 5 and Fable 5 AI models, suspending access for foreign nationals globally. This action marks a novel use of export regulations applied to AI services accessed via API, creating new legal and policy challenges.
- Export controls extended to AI models delivered via API service
- Government concerns triggered by alleged 'jailbreak' and foreign access
- Precedent complicates future AI regulation and export enforcement
What happened
On June 12, 2025, the US Department of Commerce's Bureau of Industry and Security directed Anthropic to restrict access to its Mythos 5 and Fable 5 AI models for all foreign nationals, regardless of their location. This directive was based on national security concerns and the company's inability to reliably verify user nationality.
As a result, Anthropic disabled both models entirely. The directive followed reports of a security vulnerability or 'jailbreak' of the Mythos model, raising fears that adversaries, including China-linked groups, might reverse-engineer or misuse the AI technology. This marks one of the first high-profile instances where AI services accessed via API have been subjected to export controls.
Why it matters
The significance of this move lies in the application of export controls traditionally designed for tangible goods and static software to AI models dynamically accessed as a service. Export regulations were historically built around discrete transfers, but continuous API access to AI models presents new enforcement challenges.
This action sets a precedent for how frontier AI technologies may be governed, potentially shaping the regulatory landscape and international competition in AI capabilities. It also raises questions about how governments determine when AI crosses a threshold to warrant export restrictions without transparent rulemaking or public input.
What to watch next
Governments and policymakers worldwide will likely scrutinize this approach as a test case for controlling advanced AI technologies. How export controls adapt to AI's unique characteristics—such as real-time access and global availability—will impact innovation, security, and international tech dynamics.
Additionally, the ongoing debates about defining security risks and the boundaries for AI export restrictions will shape future AI governance frameworks. Watch for public regulatory processes, legal challenges, or new legislation in response to the issues raised by this US export control enforcement.