When the US Department of Commerce sent a letter to Anthropic on June 12 requiring the company to obtain an export license before allowing any foreign person to access its Fable 5 and Mythos 5 models, the immediate effect was dramatic: Anthropic disabled both models for all customers worldwide for approximately 90 minutes while it assessed compliance options. The longer-term effect may prove more consequential, and for China’s AI industry, potentially advantageous.
The order, analyzed in depth by the Center for Strategic and International Studies, invoked the Bureau of Industry and Security’s authority under the Export Control Reform Act to restrict access to what the agency described as the most powerful AI models ever released. The letter cited national security concerns that have not been made public, though reporting has indicated that Amazon researchers identified jailbreaking methods capable of bypassing Fable 5’s safety guardrails, a finding that reportedly triggered cybersecurity concerns within the administration.
The scope of the order is sweeping. It applies not only to Anthropic’s customers but to the company’s own foreign national employees, who are required to obtain an export license before accessing the restricted models. Anthropic’s public statement acknowledged the bind directly: “The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.” As of June 16, talks between Anthropic and Trump administration officials continued without a resolution, with both sides expressing urgency but no confirmed timeline for restoring access.
The Opening for Chinese Models
CSIS’s assessment of the situation is pointed: “The inaccessibility of the most powerful AI models provides an opportunity for China to make inroads on international adoption of its models, which lag the U.S. by 7 months on average.” That gap, real but narrowing, has been the central tension in the global AI competition. American frontier models have maintained a capability lead, but Chinese open-source alternatives from DeepSeek, Alibaba’s Qwen series, and others have been closing the distance at a pace that has surprised many observers.
The Fable 5 shutdown removes the most capable option from the table for international users who cannot or will not navigate a US export licensing process. For developers in Southeast Asia, the Middle East, Africa, and Latin America, regions where Chinese AI companies have been actively cultivating adoption, the practical choice between a restricted American model and a freely accessible Chinese alternative has just become significantly easier. DeepSeek’s V4 and Alibaba’s Qwen3 series are already deployed at scale across these markets, and the disruption to Anthropic’s service creates a window for deeper entrenchment.
European politicians have drawn their own conclusions. Several have cited the Fable 5 controls as evidence for the necessity of sovereign AI capabilities, a framing that plays directly into the hands of those arguing that dependence on American AI infrastructure carries geopolitical risk. The episode has given concrete form to an abstract concern that European policymakers have been articulating for years.
The Legal and Policy Landscape
CSIS identifies three paths by which Anthropic could restore access to its models. The first is a retraction of the Commerce order after the underlying security vulnerability is resolved, the most straightforward outcome if the jailbreaking concern can be addressed technically. The second is legal action challenging the authority of BIS to impose worldwide controls on a commercial AI service, a novel legal question that has not been tested in court. The third is the implementation of identity and citizenship verification for all users, a technically complex and commercially disruptive approach that would fundamentally change how frontier AI models are distributed.
The legal authority question is not trivial. CSIS notes that the sections of the Export Control Reform Act cited by Commerce, Sections 734.13 and 744.22, are both contested as a basis for the kind of worldwide access restriction that was imposed. The Remote Access Security Act, which passed the House and would define the legal framework for controlling remote access to AI models, has not yet been enacted, leaving the current action on an uncertain legal footing.
China has already warned that US chip export bills risk disrupting global supply chains, and the broader policy context is the Trump administration’s June 2 executive order on “Promoting Advanced Artificial Intelligence Innovation and Security,” which established a voluntary framework for AI security cooperation between the government and industry. The Fable 5 order represents a shift from voluntary to mandatory, and its application to a company that has been among the most cooperative with the administration on safety issues has sent a signal to the entire AI industry about the limits of that cooperation.
For China’s AI developers, the signal is different: that the competitive landscape for global AI adoption is more open than it appeared a month ago, and that the window for establishing Chinese models as the default infrastructure for international developers may be wider than anyone anticipated.