Healthcare AI 2025

CHINA Trends and Developments Contributed by: Gil Zhang, Diana Li, Muran Sun and Huihui Li, Fangda Partners

Echoing the regulatory legislation, there is no specific regulatory authority in China responsible for the super - vision of healthcare AI. Several regulatory authorities implement regulatory responsibilities within the scope of their duties, as follows. • Medical-sector regulators, where the National Medical Products Administration (NMPA) oversees medical device registration, technical reviews and post-market surveillance for healthcare AI prod - ucts, while the National Health Commission regu - lates medical institutions and their use of AI prod - ucts and/or services. Specifically, if a healthcare AI device involves the collection, use and transfer of human genetic resource data, the National Health Commission will intervene. • Technology regulators, where CAC and MIIT are responsible for cybersecurity, AI-related regulations and data compliance. • Ancillary regulators, where the State Administra - tion for Market Regulation monitors advertising compliance, while the National Development and Reform Commission and the Ministry of Commerce oversee foreign investment. Upcoming legislation and enforcement trends Several critical legislative and regulatory initiatives moving forward in China are likely to define how healthcare AI is developed and deployed. • The Medical Device Management Law (consul - tation draft): The draft creates a single national system for managing medical-device data, pushing for interoperability and shared resources. By open - ing the door to high-quality, standardised datasets – and allowing qualified foreign clinical trial data to be used for registration in defined cases – it is expected to expedite healthcare AI innovation. • The Artificial Intelligence Law (draft listed in the State Council’s 2024 legislative plan): Although the comprehensive AI bill has been dropped from the 2025 agenda, questions of liability and risk management for medical AI remain on the table. Stakeholders should therefore keep tracking any future version of the AI Law; as more scenario- specific rules emerge, the overall compliance load will not lighten. • The Model Artificial Intelligence Law 2.0 (expert draft): Crafted by academics, this non-binding

text has yet to reach the formal legislative pipeline but may still inform later policy. Its key elements include strong backing for open-source AI through community building, explicit liability rules and new IP provisions that tackle the use of training data and personal information while clarifying protection for AI-generated output. MIIT and the NMPA run a fast-track programme that gives selected firms early regulatory and technical guidance to shorten the path to market for AI medical devices, while the National Data Administration pro - motes “regulatory sandboxes” so companies can law - fully tap clinical data. Locally, Beijing has launched the country’s first “AI data training base and pilot zone”, integrating curated datasets and compliance tools into a one-stop sandbox for LLM developers. Similar pilots are under way in Shanghai and Shenzhen. Beijing’s 2025–27 AI+Healthcare Action Plan adds expedited reviews, priority approvals and extra funding, aiming to create a globally influential, end-to-end innovation ecosystem by 2027. Administrative penalties have been levied for the use of unregistered AI-based medical software and for health data breaches, while there have been no public - ly reported cases of regulatory intervention, warnings or product recalls specifically targeting healthcare AI. Nevertheless, CAC has increased its enforcement of AI regulations since 2025. For example, in April 2025, CAC launched a three-month action plan titled “Clear and Bright Crackdown on AI Technology Abuse”, which is being implemented in two phases. This law enforcement campaign suggests that broad-sweep - ing and proactive enforcement actions and penalties are anticipated in the coming months, where: • the first phase targets six key issues, including the circulation of illegal AI products, the teaching and selling of such products, poor management of training materials, inadequate safety measures, failure to implement content labelling requirements and safety risks in critical areas; and • the second phase focuses on seven major issues, such as using AI to generate rumours, false infor - mation and pornographic content, impersonating others, engaging in online manipulation activities and infringing on the rights of minors.

CHAMBERS.COM