The United States government has abruptly reversed course on its unprecedented decision to restrict the global export of advanced artificial intelligence models. On June 30, 2026, the Bureau of Industry and Security (BIS) officially lifted the export controls it had placed on Anthropic’s frontier models, Fable 5 and Mythos 5, ending an 18-day blackout that sent shockwaves through the global tech industry.
The reversal comes after intense lobbying from the American tech sector and mounting evidence that the restrictions were inadvertently accelerating the adoption of Chinese AI models across Asia and the Global South. The episode, which industry analysts are already dubbing a “policy own goal,” highlights the delicate balance Washington must strike between protecting national security and maintaining American commercial dominance in the rapidly evolving AI landscape.
The 18-Day Blackout That Shook the AI Industry
The saga began on June 12, 2026, when the US Commerce Department took the unprecedented step of imposing export licensing requirements on Anthropic’s most capable models. The sweeping export controls required foreign users, including those in China and several allied nations, to obtain a specific license before accessing Fable 5 and Mythos 5 via application programming interfaces (APIs).
The move was widely interpreted as an attempt to prevent strategic competitors from leveraging American frontier models for military or dual-use applications. However, the immediate commercial impact was severe. Enterprise customers across Asia, suddenly cut off from their primary AI infrastructure or facing unacceptable regulatory uncertainty, scrambled to find alternatives.
Anthropic CEO Dario Amodei confirmed the lifting of the restrictions in a pointed statement. “We are pleased that the administration has recognized that broad export controls on AI models create more problems than they solve,” Amodei said. The company has since announced that Fable 5 is once again available globally, allowing enterprises to access the model via API without the burden of navigating a complex licensing regime.
Chinese AI Models Seize the Enterprise Opportunity
While Washington deliberated, Beijing’s tech giants capitalized. The 18-day window provided a unique, real-world stress test for China’s domestic AI ecosystem, and the results have rattled American policymakers.
During the ban, enterprise customers in Asia rapidly shifted their workloads to Chinese alternatives. Baidu’s ERNIE 5.0, ByteDance’s Doubao Pro, and Zhipu AI’s GLM-5.2 emerged as the primary beneficiaries of the American blackout. Data from OpenRouter showed a sharp decline in the usage share of Fable 5 and Mythos 5 across Asian markets during the restricted period, as enterprise customers routed their workloads to available alternatives.
The performance of these Chinese models surprised many international observers. TechCrunch documented how Asian AI startups rapidly launched Mythos-like alternatives during the ban, including Zhipu AI’s GLM-5.2, which gained significant enterprise traction. The rapid pivot demonstrated that Chinese AI models are winning not just on price, but increasingly on capability and reliability. For many Asian enterprises, the forced transition proved that domestic Chinese models are viable, enterprise-grade alternatives to Silicon Valley’s offerings.
Washington Recalibrates Its AI Containment Strategy
The swift lifting of the export controls suggests a significant recalibration within the Biden administration. The White House reportedly cited “concerns about competitive disadvantage for American AI companies” as a primary reason for reversing the policy.
The episode underscores a fundamental flaw in attempting to control software and weights in the same manner as physical hardware like advanced semiconductors. Unlike a lithography machine or a GPU, an API connection can be replaced almost instantly if a comparable alternative exists. By cutting off access to American models, the US government inadvertently created a massive, unearned customer-acquisition event for Chinese AI developers.
Furthermore, the ban forced international enterprises to invest the engineering hours required to integrate Chinese models into their tech stacks. Even with Anthropic’s models now available again, many of these companies may choose to maintain their new multi-model architectures, permanently reducing their reliance on American providers and diversifying their AI supply chains.
The Long-Term Impact on the US-China Tech War
The 18-day Anthropic ban will likely be remembered as a turning point in the US-China tech war. It provided empirical proof that the global AI market is no longer a unipolar American domain. When the US restricted supply, the market did not halt; it simply routed around the blockage, finding capable alternatives in Beijing and Shenzhen.
Moving forward, US policymakers face a complex dilemma. Broad export controls on AI models risk alienating international allies and driving global enterprises into the arms of Chinese competitors. However, failing to regulate the proliferation of frontier models raises legitimate national security concerns regarding their potential misuse.
For China’s AI sector, the brief window of opportunity was a resounding victory. It validated billions of dollars in domestic research and development, proved the commercial viability of their models on the international stage, and demonstrated that they are ready to fill any void left by American regulatory overreach. As the dust settles, the global AI race looks more competitive and more multipolar than ever before.