} ?>
(Yicai) Dec. 18 -- ByteDance, China’s largest unicorn, will reportedly carry out a comprehensive review in the coming days to ensure that the TikTok owner is in strict compliance with OpenAI’s terms of service following news that the ChatGPT developer has suspended its account.
ByteDance stresses compliance with OpenAI's terms of use when using its services, The Paper reported yesterday, citing a manager at the Beijing-based company. It is getting in touch with OpenAI to clarify any misunderstandings that may arise from external reports, the person added.
ByteDance pays for access to OpenAI's application programming interface, including model training and evaluation, through Microsoft to develop its large language model, US tech website The Verge reported on Dec. 15. OpenAI's model output cannot be used to develop artificial intelligence models that compete with its own products and services, according to its terms of service.
Niko Felix, a spokesperson for OpenAI, subsequently confirmed to The Verge that the San Francisco-based firm had suspended ByteDance's account. Microsoft has yet to comment.
Felix said that all API customers must adhere to OpenAi’s usage policies and while ByteDance’s use was minimal, OpenAI has blocked its account while checks are done. If ByteDance is found not to follow these policies, it will be asked to make the necessary changes or close its account, he said.
ByteDance's technical team started exploring LLMs early this year with some engineers applying the API service of OpenAI’s GPT to experimental project research on a smaller model, which was only for testing, did not provide external services, and did not plan to go to market, the manager said. The team stopped using the service after checking with ChatGPT's API terms in April, he added.
ByteDance set clear internal requirements for its LLM team in April, banning data generated by the GPT model from being added to its LLM training dataset, and trained its team of engineers to adhere to the terms of service when using GPT, the manager noted. It conducted a round of internal inspections in September and took measures to ensure compliant use of GPT's API service, he said.
Editor: Martin Kadiev