# Anthropic Ban Opens Door for Chinese Open-Source AI > Source: [https://botensten.com/articles/anthropic-ban-chinese-open-source-ai](https://botensten.com/articles/anthropic-ban-chinese-open-source-ai) (canonical) > Author: iCharles News — Botensten, https://botensten.com > Published: 2026-06-17 · Updated: 2026-06-18 ## TL;DR The US Department of Commerce ordered Anthropic to stop offering its Mythos 5 and Fable 5 frontier models to non-US nationals, including foreign employees inside the US. The move has handed a public relations boost to Chinese open-source AI labs. Shares in Knowledge Atlas Technology (Z.ai) surged over 30% after it released GLM-5.2. On OpenRouter, the top four most-used models last week all came from Chinese companies. ## What did the US government order Anthropic to do? The US Department of Commerce ordered Anthropic to stop providing access to its Mythos 5 and Fable 5 frontier models to anyone outside the United States, [according to Fortune](https://fortune.com/2026/06/16/us-anthropic-ban-open-source-ai-deepseek-zai/). The order was revealed by Anthropic on Friday, June 13. US export rules also bar the company from offering those models to any "foreign national" inside the US — including Anthropic's own employees. In response, Anthropic suspended access to both models for all users. Paul Triolo, a partner at DGA-Albright Stonebridge Group, called it "the first time that a government has ordered a model developer to restrict access to a particular model based on nationality." --- ## Which models are affected? The two models blocked are **Mythos 5** and **Fable 5** — Anthropic's frontier-tier offerings. Anthropic had previously argued Mythos was too powerful for unrestricted public release. Before the ban, Anthropic ran an early-access program called **Project Glasswing**. That program gave key institutions in about 15 countries — including US allies Japan and South Korea — controlled access to Mythos to find security vulnerabilities. --- ## Why does this matter for open-source AI? **Open-source AI models** are models that users can download and run on their own computers or cloud networks, bypassing the ability of developers or governments to restrict access. They can also be fine-tuned more easily by developers. The Anthropic ban raises the possibility that export controls could eventually hit frontier models from OpenAI or Google as well. If that happens, non-US organizations could be locked out of the best US-developed models entirely. Here's what we know so far: that scenario is already pushing governments and companies to reconsider their AI partnerships, according to Triolo. Open-source models are seen as a natural fallback, especially for governments pursuing **sovereign AI** — domestically developed and controlled AI infrastructure. --- ## Which Chinese AI companies are benefiting? Chinese labs moved quickly to claim a public relations win. Knowledge Atlas Technology — a Chinese AI lab better known as **Z.ai** — saw its shares surge over 30% in Hong Kong trading on Monday after it released **GLM-5.2**, the latest version of its open-source model. Knowledge Atlas shares are up more than 800% since their January debut. Z.ai posted on social media: "At a time when some frontier models can suddenly become unavailable, we choose to believe in a different path." The company added that "frontier intelligence should not belong to only a few people, nor be subject to withdrawal by a handful of rules at any moment," in a clear reference to the Anthropic news, [as reported by the South China Morning Post and cited by Fortune](https://fortune.com/2026/06/16/us-anthropic-ban-open-source-ai-deepseek-zai/). --- ## What does OpenRouter usage data show? **OpenRouter** is a popular platform for accessing different AI models. Last week, the top four most-used models on OpenRouter all came from Chinese companies: **[DeepSeek](/articles/deepseek-entity-list-blacklist-delayed)**, **MiniMax**, **Tencent**, and **Xiaomi**. Chinese open-source models have proven popular not just inside China but also across many developing countries, where they are viewed as a strong price-to-performance option. --- ## How does this connect to China's broader tech strategy? Neil Shah, vice president of research at Counterpoint Research, said the US ban is "a great move for China." He noted that while China is not at the cutting edge due to chip export controls, "they have their own silicon and their own software." The US placed controls on advanced chip and chipmaking equipment sales to China in 2022 under the Biden administration. That move accelerated China's push toward tech self-sufficiency — a trend the Anthropic ban now appears to reinforce. The [US stance on DeepSeek](/articles/deepseek-entity-list-blacklist-delayed) has been a separate but related flashpoint in that broader competition. --- ## Who else is watching this closely? Triolo said companies and governments "will start reconsidering how they are approaching application development based on a particular model." For governments, the question is which companies they want to partner with for [sovereign AI](/articles/altman-amodei-hassabis-g7-summit) deployments. The Anthropic case is the first known instance of a government ordering a model developer to restrict access by nationality. That precedent has implications for the entire [AI productivity](/articles/ai-productivity-deficit-yale-budget) and enterprise software stack that non-US organizations have built on US models. --- ## Key facts at a glance | Detail | Fact | |---|---| | Models blocked | Mythos 5, Fable 5 | | Order issued by | US Department of Commerce | | Announced by Anthropic | Friday, June 13, 2026 | | Z.ai (Knowledge Atlas) share surge | +30% on Monday | | Z.ai shares since January debut | +800% | | Top 4 models on OpenRouter last week | DeepSeek, MiniMax, Tencent, Xiaomi | | Project Glasswing countries | ~15, including Japan and South Korea | --- The most important confirmed fact: as of June 13, Anthropic has suspended access to Mythos 5 and Fable 5 for all users globally in response to the US Department of Commerce order. ## Frequently asked questions **What did the US government order Anthropic to do?** The US Department of Commerce ordered Anthropic to stop providing access to its Mythos 5 and Fable 5 models to anyone outside the United States. US export rules also bar Anthropic from offering the models to foreign nationals inside the US, including its own employees. Anthropic announced the order on Friday, June 13, 2026, and suspended access to both models for all users. **What is Project Glasswing?** Project Glasswing was Anthropic's early-access program for its Mythos model. It gave key institutions in about 15 countries — including US allies Japan and South Korea — controlled access to Mythos to help uncover security vulnerabilities. Anthropic had argued Mythos was too powerful for unrestricted public release, which is why the program used a limited, vetted rollout. **Why did Z.ai (Knowledge Atlas Technology) shares surge 30%?** Shares in Knowledge Atlas Technology, the Chinese AI lab known as Z.ai, surged over 30% in Hong Kong trading on Monday after it released GLM-5.2, the latest version of its open-source model. The timing coincided with the Anthropic ban controversy. Z.ai publicly referenced the news, saying frontier intelligence should not be subject to withdrawal by a handful of rules. **Which Chinese AI models are most used on OpenRouter right now?** As of last week, the top four most-used models on OpenRouter — a popular platform for accessing AI models — all came from Chinese companies: DeepSeek, MiniMax, Tencent, and Xiaomi. Chinese open-source models have gained traction not just in China but also across many developing countries, where their price-to-performance ratio is seen as a strong advantage. **Is this the first time a government has restricted an AI model by nationality?** Yes, according to Paul Triolo, a partner at DGA-Albright Stonebridge Group. He described the US Department of Commerce order against Anthropic as "the first time that a government has ordered a model developer to restrict access to a particular model based on nationality." He said companies and governments will now reconsider how they approach AI application development and sovereign AI partnerships.