CHINA Trends and Developments Contributed by: Vincent Wang, Xinyao Zhao and Amy Cao, Global Law Office
been registered successfully with CAC as of 31 December 2024; 105 services that integrate third-party AI interfaces were also registered. The registration process for these integrated services is simpler than the standard registra - tion process. National standards regarding the safety of AI services In 2024, China issued the national standards for AI service security, showcasing the best compli - ance practices in generative AI business opera - tion. These standards also serve as references for authorities in interpreting and enforcing relat - ed regulations. • In March 2024, National Technical Commit - tee 260 on Cybersecurity of Standardisation Administration of China (TC260) released a standard titled “Basic Security Require - ments for Generative AI Services (TC260- 003)”, building on the Interim Measures on Generative AI Services. It provides detailed guidelines for processing training data, model training and service provision, ensuring that generative AI brings user convenience without triggering security risks. This stand - ard serves as a key reference for companies seeking registration with CAC for their gen - erative AI services. • In the same month, a draft standard titled “Cybersecurity technology – Generative artificial intelligence data annotation security specification” was introduced and opened for public comments. This standard addresses the safety of manual data annotation used for training models and supports the imple - mentation of data annotation provisions in the Interim Measures on Generative AI Services. • In the following months, two additional recommended national standards for genera - tive AI were released for public comments:
“Cybersecurity technology – Security speci - fication for generative artificial intelligence pre-training and fine-tuning data”; and “Cybersecurity technology – Basic security requirements for generative artificial intel - ligence service”. Notably, in 2025, a compul - sory national standard on generative AI was also released: “Cybersecurity technology – Labelling method for content generated by artificial intelligence”. This type of compulsory standard is uncommon in this field, and thus warrants close attention for companies in the data labelling industry. Highlights of data assets and use Data as an asset On 1 January 2024, the Interim Provisions on Accounting Treatment for Enterprise Data Resources came into effect. These provisions allow data to be recorded as an asset on enter - prises’ balance sheets and clearly define the scope of data resources and applicable stand - ards. As of the end of August 2024, public statistics show that 41 companies listed on the stock exchanges in China have included data assets on their balance sheets, totalling approximately CNY1.3 billion and potentially boosting their revenues. These companies are primarily in the information transmission, software and IT servic - es, and manufacturing sectors. For example, an AI company with significant intellectual property and data resources embedded in its AI models lacks tangible collateral and aims to leverage these data resources as an asset for financing to address capital constraints. Despite this, companies still face uncertainties and challenges in listing data as an asset. Key issues include confirming data asset rights, pro - tecting individual data security and privacy, and
94
CHAMBERS.COM
Powered by FlippingBook