Following OpenAI's lead, Anthropic has announced restrictions on Chinese-controlled enterprises using its AI services, raising questions about whether this development creates opportunities for Chinese AI model companies.

**US AI Companies Close Doors to Chinese Enterprises**

Today, US AI company Anthropic published an update on its website regarding "updated sales restrictions for unsupported regions," announcing the immediate cessation of services to groups and subsidiaries where majority ownership is held by Chinese capital.

According to Anthropic's new policy statement, any enterprise directly or indirectly controlled by Chinese entities with ownership exceeding 50% is prohibited from using Anthropic services. This restriction covers not only mainland Chinese companies but also extends to overseas subsidiaries, cloud service intermediaries, and organizations with Chinese-backed investment entities. Anthropic stated this measure addresses legal, regulatory, and national security risks.

Anthropic's flagship product is the Claude series of large language models, currently popular among developers in the industry for their programming capabilities. Claude series APIs are sold both directly by the company and integrated into cloud service platforms from Google, Amazon, Microsoft, and others to provide enterprise services.

Anthropic's policy is strict, meaning any company directly or indirectly controlled by Chinese entities (with ownership exceeding 50%), regardless of its operational location, can no longer use Claude or other Anthropic AI services, nor access these models through related enterprise cloud services.

This isn't the first time a US AI company has imposed restrictions on Chinese-controlled enterprises. Besides Claude, nearly all prominent US AI large language models prohibit use by Chinese companies or users.

In mid-June 2024, OpenAI announced that retired US Army General and former National Security Agency (NSA) Director Paul M. Nakasone had joined its board to provide cybersecurity advice. Subsequently, on June 25, multiple users received push notifications from OpenAI stating that API (Application Programming Interface) services from non-supported countries and regions would be blocked starting July 9, 2024. This meant OpenAI officially terminated API service access to mainland China and Hong Kong from July 9, 2024. At that time, numerous domestic companies quickly launched migration plans to attract OpenAI developers.

**Preventing China's Advanced AI Development**

Anthropic's statement also declared it would continue advocating for stronger US export controls to prevent adversaries from developing advanced AI capabilities.

Anthropic was founded in 2021 by siblings Dario Amodei and Daniela Amodei. Dario Amodei holds degrees from top institutions including Stanford and Princeton, initially focusing on physics before transitioning to neuroscience and AI. Early in his career, he worked at Baidu's Silicon Valley AI lab as a researcher under Andrew Ng, later moving to Google as a senior researcher. In 2016, he joined OpenAI, overseeing company safety and leading teams developing GPT-2 and GPT-3.

According to foreign media reports, Anthropic CEO Dario Amodei has consistently advocated publicly for technology sanctions against China to prevent adversaries from developing advanced AI capabilities.

Anthropic's flagship products include the Claude series of large language models, encompassing Claude 3, Claude 3.5 Sonnet, and Claude 4.1, which excel in code generation, multimodal interaction, and long-context processing. Unlike OpenAI, which serves both consumer and business markets, Anthropic focuses more on providing AI solutions for enterprise clients. Its programming assistance tool Claude Code has experienced rapid growth, contributing substantial revenue.

On September 3, Anthropic announced completion of a $13 billion Series F funding round, reaching a valuation of $183 billion and becoming the world's third-largest AI unicorn, trailing only ByteDance and OpenAI. According to data revealed in the funding announcement, as of August 2025, Anthropic's annualized revenue exceeded $5 billion with over 300,000 enterprise clients.

Since 2021, Anthropic has completed 8 funding rounds, raising over $17 billion total. Its investor lineup is impressive, including Amazon, Google, Fidelity Management & Research Company, Lightspeed Venture Partners, as well as sovereign wealth funds like Qatar Investment Authority and Singapore's Government Investment Corporation (GIC).

**Opportunity or Challenge for Domestic Large Language Models?**

However, some industry insiders believe that as the technological iron curtain in artificial intelligence gradually descends, this actually provides development opportunities for Chinese large language model companies.

According to a Gartner report, driven by open models, frugal AI, and engineering capabilities, China's generative AI production deployment rate is soaring: jumping from 8% in 2024 to 43% in 2025. International market blockades may actually accelerate the development pace of China's large language model industry. Like in the semiconductor field, external pressure is driving China's AI industry toward a path of independent innovation.

In fact, China's large language model industry already possesses considerable strength. Gartner's "2025 China AI Trends" report notes that DeepSeek's open-source release at the beginning of the year was a landmark event, leading the global open-source wave with excellent performance and profoundly changing market dynamics.

The gap between open-source and closed-source models is rapidly narrowing. Past advantages of stacking chips and competing on scale are weakening, as open-source models perform similarly with lower inference costs. In tasks like text generation, web development, and Copilot functions, open-source models perform excellently.

From a global perspective, China has abundant AI talent reserves. The proportion of papers by Chinese authors at top AI conferences jumped from 29% in 2019 to 47% in 2022.

Industry professionals recommend that developers or enterprises previously relying on Claude's API consider the following directions: actively monitor and trial mainstream domestic large language models, which are rapidly improving and typically provide documentation and support better suited to domestic developer needs. They can also research and deploy some international open-source large language models, such as the LLaMA series and Mistral. However, attention should be paid to complying with corresponding open-source agreements and data compliance issues.