AsianFin--Chinese artificial intelligence (AI) companies are scrambling to attract users of OpenAI's technology in response to emails sent by the U.S. firm regarding their decision to restrict access to its application programming interface (API) in China and other countries. This API is crucial for developers looking to integrate OpenAI's AI models into their own products.
Since late Monday, Chinese users of the platform have received emails warning they are in a "region that OpenAI does not currently support" and that additional measures to block API traffic from unsupported regions would be taken starting July 9.
Although OpenAI has made its AI services available in over 160 countries and regions, these products remain officially inaccessible in Chinese mainland. Users in these regions have been resorting to virtual private networks (VPNs) or third-party applications to access ChatGPT. Meanwhile, developers have been using proxies and outbound servers to circumvent restrictions.
OpenAI's move could impact Chinese companies developing their own services based on OpenAI’s large language models (LLMs), the technology that powers ChatGPT and similar generative AI applications, according to some experts.
In response, Zhipu AI, one of China’s top generative AI start-ups, said it would help affected developers transfer to its platform.
“We are rolling out a special migration plan for shifting to home-grown LLMs,” the Beijing-based start-up said on Tuesday, offering perks such as 150 million free tokens, as well as tailored training courses to support such moves.
Baidu, China's leading AI developer, said it would launch an "inclusive program" offering new users free migration to its Ernie platform.
For OpenAI users, Baidu will provide additional Ernie 3.5 flagship model tokens, matching the scale of their OpenAI usage, Baidu's cloud unit said in a statement. Tokens are units of text processed by AI models.
Alibaba Cloud also joined in, offering free tokens and migration services for OpenAI API users through its AI platform. The company's Qwen-plus model is priced significantly lower than GPT-4, according to Alibaba.
Silicon Cloud, a Chinese AI infrastructure (AI Infra) company, announced a "Free Token" plan on June 25, offering top open-source models like Qwen2-7B, GLM-4-9B, and Yi-1.5-9B for free.
On Tusday evening, Ling Yi Wan Wu, an AI model unicorn founded by Kai-Fu Lee, announced the "Yi API Two-Fold Alternative Plan" for OpenAI users. This plan offers smooth migration to the Yi series models, providing high-performance and cost-effective alternatives corresponding to various OpenAI models.
MiniMax, a leading general artificial intelligence tech company, launched a zero-cost nanny-style migration plan, with an open platform interface fully compatible with OpenAI, allowing one-click transfer.
Earlier, MoonShot AI revealed that to facilitate quick access to model capabilities for AI developers, MoonShot AI's open platform API is compatible with OpenAI. Developers can transfer smoothly without any additional modifications beyond basic parameters to immediately experience the capabilities of MoonShot models.
Many Chinese companies have released chatbots powered by their own AI models over the past year.
According to China Fund, the main channels for using OpenAI technology in China offer undifferentiated API capabilities: first, connecting to OpenAI's official API; second, connecting to OpenAI technology provided by Microsoft Azure (the only compliant channel in China, requiring corporate qualifications to apply).
China Fund suggests that while OpenAI's cessation of API services in China may cause short-term disruptions, it could serve as a catalyst for domestic AI technology innovation and research in the long run. It may prompt Chinese AI companies to accelerate independent research and development, advancing the growth and innovation of local AI technology. Domestic developers and enterprises might increasingly turn to domestic models to mitigate future risks.
