US Export Controls on Anthropic Models Trigger EU Push for Tech Sovereignty
Key Takeaways
- ▸US government issued export control directive requiring Anthropic to restrict Mythos 5 and Fable 5 access to US citizens only, effectively making the models unavailable globally
- ▸Directive allegedly triggered by a reported jailbreak vulnerability shared by Amazon CEO Andy Jassy to US Treasury and other officials
- ▸EU Commission views incident as validation of its Technological Sovereignty Package and argument that Europe cannot depend on US-controlled AI infrastructure
Summary
The US government issued an export control directive requiring Anthropic to restrict access to its Mythos 5 and Fable 5 cybersecurity models exclusively to US citizens, effectively taking the advanced models offline globally since citizenship cannot be verified online. The order also prevents some Anthropic staff from accessing their own models.
The directive allegedly stemmed from concerns about a potential jailbreak vulnerability in the models, with Amazon CEO Andy Jassy reportedly flagging the issue to US Treasury Secretary Scott Bessent and other officials. Anthropic said the government shared only verbal evidence of a narrow jailbreak involving requesting the model to analyze a specific codebase and fix security flaws. Anthropic executives are scheduled to meet White House officials this week to seek clarification on the government's specific concerns. The European Commission responded by highlighting the incident as evidence that Europe must strengthen its technological sovereignty, arguing that foreign governments should not be able to unilaterally restrict access to technology that allied partners rely on.
- The practical impossibility of verifying user citizenship online meant the security measure resulted in a complete service blackout rather than targeted restrictions
