US Export Controls on Anthropic Boost Open-Source AI, Chinese Models Gain Traction
The U.S. government has ordered Anthropic to cease providing its Mythos and Fable 5 AI models to non-U.S. nationals, leading the company to suspend access for all users. This decision is reportedly providing a significant boost to the adoption of open-source AI models, particularly those developed by Chinese AI labs such as DeepSeek and z.ai (Knowledge Atlas). Chinese companies are leveraging the controversy, with Knowledge Atlas's shares surging after the release of its GLM-5.2 model. Demand for Chinese AI models has reportedly surpassed U.S. models on platforms like OpenRouter, highlighting a potential shift towards tech self-sufficiency for non-U.S. entities.
The U.S. government's decision to stop Anthropic from offering its Mythos and Fable 5 models to non-U.S. nationals is expected to significantly increase the adoption of open-source AI models, including those from Chinese AI labs like DeepSeek and Moonshot AI.
Open-source models allow users to download and run them on their own systems, bypassing controls from developers and governments. These models also offer greater flexibility for developers to fine-tune them for specific applications.
Chinese AI developers are already seeing benefits from the Anthropic situation. Knowledge Atlas, also known as z.ai, experienced a more than 30% surge in its Hong Kong-traded shares on Monday following the release of its latest open-source model, GLM-5.2. The company stated on social media 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 what was seen as a direct reference to the Anthropic news.
Reports indicate that demand for Chinese models has surpassed that for U.S. models on OpenRouter, a platform for accessing various AI models. Last week, the top four most-used models on the platform reportedly originated from Chinese companies: DeepSeek, MiniMax, Tencent, and Xiaomi. These models are gaining popularity not only in China but also in developing countries due to their balance of price and performance.
This U.S. action may also support China's broader objective of tech self-sufficiency, which gained momentum after the Biden Administration implemented controls on advanced chip sales in 2022. Neil Shah, vice president of research at Counterpoint Research, noted that while China may not be at the cutting edge due to export controls, it possesses its own silicon and software.
Anthropic confirmed on Friday that the U.S. Department of Commerce had ordered it to restrict access to its frontier models for anyone outside the U.S. The interpretation of U.S. export rules also meant the company could not offer these models to any "foreign national" within the U.S., including its own employees. As a result, Anthropic suspended access to these models for all users. Previously, Anthropic had initiated an early-access program called Project Glasswing for key institutions in about 15 countries to use its Mythos model and identify security vulnerabilities.
Experts suggest that the U.S.'s move against Anthropic raises concerns that frontier models from other labs, such as OpenAI or Google, could also face similar export controls. This could lead to non-U.S. organizations being completely barred from accessing leading U.S.-developed models. Open-source models could serve as an alternative, especially for governments interested in developing sovereign AI. Paul Triolo, a partner at DGA-Albright Stonebridge Group, highlighted that this is the first instance of a government ordering a model developer to restrict access based on nationality, prompting companies and governments to reevaluate their application development strategies. (Source: Fortune)
