Chinese AI startups confront challenges as OpenAI ends API services in China

Chinese startups that had integrated OpenAI received troubling news on 25 June. In an official email, OpenAI stated, “Our data shows that your organisation is receiving API traffic from regions that OpenAI does not currently support.” The email further mentioned that starting from 9 July, OpenAI will begin blocking API traffic from unsupported countries and regions. It is reported that OpenAI’s API is currently available in 161 countries and regions, but China is not included. This effectively means OpenAI has announced the termination of its API services in China.

To understand APIs in layman’s terms, let’s consider an example from Amazon. APIs are mechanisms that enable two software components to communicate using a set of definitions and protocols. For instance, the weather bureau’s software system contains daily weather data. The weather app on your phone “talks” to this system via APIs, showing you daily weather updates.

In March 2023, OpenAI began offering API services, allowing users to access a variety of models, including GPT-4o. Previously, OpenAI’s ChatGPT had strict restrictions, preventing users in unsupported regions from registering. However, OpenAI’s API services were less restrictive, enabling some developers to integrate the OpenAI API and provide services, thus allowing them to create substantial models or AI applications.

The reason for OpenAI’s sudden termination of API services to certain regions may be linked to recent developments. On 13 June OpenAI welcomed a new board member, retired U.S. Army General Paul M. Nakasone, a top expert in cybersecurity and global cyber defence. Paul played a key role in establishing the U.S. Cyber Command and served as the Director of the National Security Agency. He joined OpenAI to advise on key safety and security decisions across all OpenAI projects and operations.

Furthermore, on 22 June, the U.S. Treasury Department issued a draft rule proposing regulation of certain U.S. investments in semiconductors, microelectronics, quantum computing, and AI. The rule also suggested prohibiting or restricting Chinese investments in AI and other technologies. OpenAI’s move is seen as a response to a combination of factors, including legal compliance and data security.

The termination of OpenAI’s API service to China will have significant consequences. On one hand, it will deal a severe blow to companies that rely on OpenAI’s Large Language Model (LLM), potentially eliminating those without core technology and competitive strength if they fail to find alternative solutions. On the other hand, it will push Chinese AI companies to accelerate their independent research and development and encourage more startups to choose domestic models to avoid future risks.

After the news spread, domestic large model vendors swiftly moved to capture OpenAI’s market share by announcing user migration programmes. Zhipu AI unveiled “a special moving plan for OpenAI API users” to help users switch to a domestic LLM. Specifically, Zhipu will provide developers with 150 million tokens and a series of migration training sessions from OpenAI to GLM. Baidu Cloud Qianfan introduced a LLM universal plan, offering free usage, training, and migration services for newly registered enterprise users, with the flagship Wenxin model available for free for the first time. Additionally, Aliyun announced that it would offer OpenAI API users the most cost-effective Chinese large model alternative, providing 22 million free tokens and exclusive migration services for Chinese developers.

Despite these efforts, OpenAI still holds irreplaceable advantages, particularly in reasoning and data processing, which are challenging to replicate domestically. Although OpenAI has terminated its API service in China, it remains to be seen whether enterprise users can still indirectly utilise OpenAI’s technical support through cooperation with the Microsoft Azure platform.


Join our newsletter