Zhipu AI ramps up efforts for data privacy

Chinese artificial intelligence startup Zhipu AI, or Z.ai in overseas markets, will introduce a "no data retention" option on its model-as-a-service platform, which marks one of the strictest data privacy arrangements offered by a major Chinese large language model provider.

Under the new service arrangement, inputs sent by companies and developers to Zhipu's models, as well as the models' outputs, will not be stored after a request has been processed, the Beijing-based firm said.

Domestic AI companies increasingly compete not only on model intelligence and price, but also on whether customers can trust them with sensitive information. As generative AI moves beyond consumer chatbots into banks, factories and tech companies, models are increasingly being asked to process contracts, customer information, internal reports, source code and other proprietary material.

Whether that information is stored, reused to train models or potentially leaked and exposed has therefore emerged as a key consideration for businesses deciding whether to deploy generative AI across their operations.

For an ordinary chatbot user, the fate of a casual prompt may be a secondary consideration. For a bank processing client information or a tech company feeding proprietary source code into an AI coding assistant, data retention can determine whether a product is allowed inside the organization at all.

Xiao Rongmei, deputy director of the Policy and Economics Research Institute, China Academy of Information and Communications Technology, previously warned that information entered into generative AI systems may contain personal information and commercial secrets.

If such information is subsequently used in model training and reproduced in generated content, it could create risks of data leakage, Xiao said, making protection of user inputs and usage records an important part of AI governance.

China has required generative AI services providers to protect users' input information and usage records, and comply with requirements governing the handling of personal information.

Yu Xiaohui, president of the academy, said at a separate conference that AI governance and industrial competitiveness are increasingly intertwined, with stronger governance helping to open up adoption in high-value sectors including finance and manufacturing.