Chinese tech giant ByteDance has confirmed its “very limited” use of OpenAI’s technology in developing its own AI models. This revelation follows OpenAI’s suspension of ByteDance’s account due to concerns about GPT data usage. A ByteDance spokesperson stated that the company’s engineering team employs OpenAI’s GPT, alongside other third-party models, for evaluation and testing purposes. ByteDance holds a license from Microsoft for GPT APIs, utilizing them for products and features outside of China, where OpenAI’s services are unavailable.
The Beijing-based company, owner of TikTok, asserts that it solely relies on its internally developed AI model for Doubao, a ChatGPT-like tool exclusively available in China. Recent reports suggested Byte Dance’s alleged “secret” use of OpenAI’s technology in training its large language model (LLM), raising concerns of a violation of OpenAI’s terms of service.
In response, OpenAI confirmed the suspension of ByteDance’s account, emphasizing adherence to usage policies. ByteDance maintains that its use of the OpenAI API was minimal and that they are cooperating with the investigation. The Verge, citing internal documents, claimed Byte Dance’s reliance on OpenAI API throughout the development of its foundational LLM, known as Project Seed.
Byte Dance informed that, in April, internal guidelines were established to prevent the inclusion of GPT-produced data in the training datasets of its self-developed models. Compliance training on GPT’s terms of service was also provided to the engineering team. The unfolding situation adds a layer of complexity to the already existing US-China rivalry in emerging technologies.