Following a US export control directive issued in June 2026, Anthropic has disabled access to its flagship AI models Fable 5 and Mythos 5 for all foreign nationals worldwide, including foreign employees. This significant move highlights tensions between AI safety enforcement and broader geopolitical considerations in technology governance.
- Anthropic barred from providing Fable 5 and Mythos 5 to non-US nationals worldwide.
- US export controls target alleged jailbreaking vulnerabilities in Anthropic’s models.
- Anthropic counters that similar vulnerabilities exist in other major AI models like OpenAI’s GPT-5.5.
What happened
The US government issued an export control directive on June 12, 2026, requiring Anthropic to restrict access to its AI models Fable 5 and Mythos 5 for all foreign nationals, regardless of their location. Anthropic responded by disabling these models globally for any user considered a non-US national, including its own foreign-national employees. The directive does not affect access to Anthropic’s other AI models.
Anthropic received only verbal notification of the government's concerns, which cited the possibility of a narrow, non-universal jailbreak method that could bypass some safety features in Fable 5. The company describes this vulnerability as minor and previously known, arguing that nothing on this scale affects the model universally or justifies a global suspension.
Why it matters
This development underscores growing US attempts to control the international dissemination of advanced AI technology under national security pretexts. By invoking export controls, the US may be setting new precedents about which AI capabilities are permissible to share globally, especially amid rising AI risks and geopolitical competition.
Anthropic’s pushback reveals tensions within the AI industry regarding security vulnerabilities and regulatory boundaries. The company highlights that similar jailbreak capabilities can be found in other widely used models, such as OpenAI's GPT-5.5, calling into question whether this specific incident justifies a sweeping restriction. This case also shows the challenge of balancing robust AI safety with innovation and international access.
What to watch next
Attention will turn to how other leading AI providers respond to this export control regime and whether they face similar restrictions or develop new compliance strategies. The US government's enforcement approach might influence global AI governance frameworks and cross-border technology flows.
Observers will also watch for developments in Anthropic’s legal and diplomatic efforts to challenge or negotiate this directive, as well as any clarifications regarding the precise nature of the alleged security threat. Finally, monitoring industry responses to improve model safeguards or call for clearer regulatory standards will be critical as AI capabilities advance.