Navigating China’s Evolving AI Regulation: Advice for MNCs

A key part of China’s regulatory framework for artificial intelligence requires generative AI providers to pass the large model security assessment and register their algorithms with the Cyberspace Administration of China (CAC). Such supervision has led to significant progress, with 940 deep synthesis services successfully completing algorithmic filing and 114 generative AI services passing large model security assessments by April 2024. Also, the courts of China have decided on several landmark AI cases, particularly following the implementation of the Interim Measures for the Management of Generative Artificial Intelligence Services (“Gen AI Measures”). This article thus aims to provide advice for multinational corporations (MNCs) seeking to offer or use AI services in China in light of such.

China’s legislation on AI governance focuses on four major aspects: algorithm compliance, content management, data protection, and copyright protection. While the regulatory requirements on the above matters may scatter in multiple interim measures, an omnibus AI Law has been included in the legislative plan of the State Council and will be issued in the forthcoming years to establish a more comprehensive AI regulatory system.

While the AI regulatory framework evolves, China has also witnessed some noteworthy cases concerning AI-generated content.

In the first AI platform infringement case, a website provides image-generation services and users can generate Ultraman (a Japanese anime character) images based on related keywords. Ultraman’s copyright holder in China therefore claimed that website A infringed its right to reproduce and adapt Ultraman’s works. Although the website argued that the image-generation service was provided by a third party, the court found that it still constitutes an AI service provider as defined in the Gen AI Measures and should fulfill the corresponding obligations. Therefore, according to Article 14 of the Gen AI Measures, when the website discovers illegal content, it shall promptly take rectification measures. Specifically, the website is ordered to conduct keyword filtering to prevent its services from continuing to generate images that are substantially similar to Ultraman’s works. However, since the model underlying the image-generation service is provided by a third party, the court did not support the copyright holder’s claim that the website shall delete the Ultraman materials from the training dataset.

Another case is the first AI voice personality rights infringement case. In this case, the plaintiff, a voice actor, was commissioned by a company for sound recording and agreed to assign to the company the respective copyrights. This company thereafter licensed the sound recording to a third party as AI training material. Although both the company and the third party argued that they had legal rights to utilise the plaintiff’s sound recordings, the court held that the copyright in sound recordings does not include the plaintiff’s personality rights over his voice, such that the reproduction or mimicking of the plaintiff’s sound in the accused AI service constitutes infringement.

These cases highlight China’s heightened scrutiny over the legality of AI-generated content. Combining China’s current AI regulatory framework with the emerging AI cases, MNCs aiming to provide or use AI services in China should consider the following compliance advice.

For MNCs providing AI services:

For MNCs using AI services:

While the regulatory landscape surrounding AI in China is evolving, entities should proactively ensure compliance with existing laws and regulations via algorithms compliance, content management, data protection, and copyright protection. Furthermore, entities should keep informed of the future introduction of the comprehensive AI Law to stay ahead of the curve and minimise the non-compliance risks.