China rejects OpenAI’s GPT-5 trademark bid, here's why

ChatGPT

OpenAI’s attempt to register “GPT-5” as a trademark in mainland China has been rejected by the country’s Trademark Office, under the China National Intellectual Property Administration. The filing, made through subsidiary OpenAI OpCo, is now pending appeal, according to a report by South China Morning Post.

The decision adds to a series of rejections for the San Francisco-based AI company. Between March and November 2023, Chinese regulators denied applications to trademark “ChatGPT” and various GPT model names, including GPT-4, GPT-5, GPT-6, and GPT-7. All remain under appeal, as per the report.

The refusal mirrors OpenAI’s struggles elsewhere. In February 2024, the US Patent and Trademark Office turned down applications for “ChatGPT” and “GPT,” ruling that the terms merely describe features or functions of the products, rather than serving as distinctive brand identifiers.

In China, the descriptive nature of “GPT” — shorthand for generative pre-trained transformer, a widely used type of large language model — makes it particularly difficult to secure as a trademark, according to Beijing-based corporate IP expert Wu Zheng. GPT models are foundational in training generative AI chatbots, and the acronym is already broadly used in the industry.

Despite Microsoft-backed OpenAI’s rollout of AI services in over 160 countries, its products remain officially unavailable in mainland China and Hong Kong. Access is possible only through VPNs, third-party applications, or proxy servers for developers.

OpenAI unveiled GPT-5 last week, calling it its “smartest, fastest, most useful model yet,” with CEO Sam Altman describing it as akin to “a PhD-level expert in anything.” The company sees the model as a major step toward embedding AI intelligence into core business operations — but in China, at least, it will have to fight longer to secure the name.