① The capacity of grassroots institutions, the cognitive disparity among patients, and the responsiveness of data and regulatory systems collectively determine the vitality of AI medical technology. ② iFlytek Chairman Liu Qingfeng pointed out to a Cailian Press reporter that services such as AI health profiling and intelligent follow-up should be incorporated into the national basic public health service system. ③JD.com Technical Committee Chairman Cao Peng proposed promoting an AI-driven closed-loop model of “medical-testing-diagnosis-pharmacy.”
Cailian Press, March 11 (Wang Junxian and Lu Afeng) As artificial intelligence gradually moves from the laboratory to real-world clinical scenarios, its role is evolving from a single auxiliary tool to an infrastructure component of the health service system, subtly reshaping the operational logic of the health system.
As applications dig deeper, the industry increasingly recognizes that the real challenges lie far beyond algorithms. The capacity of grassroots institutions, cognitive disparities among patients, and the responsiveness of health data and regulatory frameworks all collectively determine the feasibility of AI medical technologies.
A series of detailed proposals submitted by representatives of the medical field and committee members during the two national sessions in 2026 marked the transition of the field from the “technological exploration stage” to the “applied implementation stage”. Liu Qingfeng, deputy member of the National People’s Congress and chairman of iFlytek (002230.SZ), pointed out to a Cailian Press reporter that primary healthcare faces practical challenges such as limited human resources and lack of precision in chronic disease management. He specifically recommended that services such as AI health profiling and intelligent follow-up be incorporated into the nation’s basic public health service system. Cao Peng, a member of the National Committee of the Chinese People’s Political Consultative Conference and chairman of the JD.O. Technical Committee, also proposed promoting a closed-loop model of “examination, diagnosis, and pharmacy” using AI. Dai Lizhong, deputy member of the National People’s Congress and chairman of Saint Biotechnology (688289.SH), provided insights and suggestions on the main bottlenecks in the commercialization of AI in healthcare.
There is an urgent need to bring AI healthcare to the grassroots level.
In China’s healthcare system, primary healthcare institutions have long been responsible for managing chronic diseases. However, as the population ages, the pressure to manage chronic diseases continues to increase.
According to data from the National Health Commission and related epidemiological studies, the number of hypertension patients in China is over 300 million, and the number of diabetes patients is about 140 million. Meanwhile, according to data from the National Bureau of Statistics, by the end of 2025, China’s population aged 60 and above will reach approximately 323 million people, and the demand for chronic disease management is expected to continue to increase.
Against this background, primary medical institutions are required to take on multifaceted responsibilities, including not only basic diagnosis and treatment, but also follow-up for chronic diseases and health management. The problems of limited manpower and insufficient accuracy are becoming increasingly prominent.
Liu, vice member of the National People’s Congress “Facing new situations such as the aging of the population and the increasing burden of chronic diseases, problems at the grassroots level continue, such as insufficient human resources, lack of precision in chronic disease management, and lack of personalized health education. Practices in various regions have proven that AI technology can effectively help family doctors optimize contract performance management, achieve accurate chronic disease management, and provide personalized medical services to residents,” Qingfeng told reporters.
To address these issues, he specifically recommended that services such as AI health profiling, intelligent follow-up, health consultation, and risk assessment be incorporated into the country’s catalog of basic public health services and that service standards and target populations be defined. He also suggested that the performance of AI applications be linked with grassroots evaluation mechanisms to encourage active promotion and improve service quality.
To transform this auxiliary function into systematic service efficiency, Cao Peng proposed a more practical “medical-test-diagnostic-medicine” closed-loop model. He encouraged Internet medical platforms to develop an AI-driven comprehensive medical service loop and proposed to support the integrated model innovation of “AI triage – online consultation – home examination – medicine delivery – rehabilitation management”. Chronic disease management and health management services that utilize AI that meet the standards should be included in the scope of primary care physician contract services, and a pilot program for medical insurance payment should be implemented. At the same time, we will implement a special action plan to strengthen the inclusiveness of AI healthcare at the grassroots level by creating customized AI entities for local doctors and rural practitioners, promoting the “AI digital doctor” model, and comprehensively strengthening primary level diagnostic and treatment capabilities.
However, technology adoption still requires institutional incentives. Dai Lizhong pointed out that AI healthcare is not well integrated into performance evaluation and reward systems, leading to a lack of incentive mechanisms and insufficient participation enthusiasm among grassroots medical institutions and staff. He recommended incorporating AI healthcare services into medical institutions’ performance evaluation systems and defining quantifiable service indicators to guide medical institutions to optimize service processes and improve service quality.
Solving the key bottlenecks of AI healthcare commercialization through data
The large-scale application of AI in healthcare relies on the provision of high-quality data as its core driver.
However, Dai Lizhong pointed out that medical data is characterized by privacy and fragmentation. Currently, China’s medical data disclosure and sharing mechanisms remain underdeveloped, and AI healthcare companies cannot effectively access most anonymized medical data. As a result, there is a lack of high-quality and large-scale training datasets, hindering the development of high-precision AI models tailored to China’s medical scenarios. In this way, technological innovation has fallen into the dilemma of not being able to cook the rice, which has become a central bottleneck preventing the commercial implementation of AI in healthcare.
To address this issue, he proposed accelerating the release of anonymized medical data and solidifying the “digital foundation” of AI, using “secure controls and open sharing” as a core approach to solving data bottlenecks. First, based on the “15th Five-Year Plan” national health data platform, we will integrate resources such as electronic health records, electronic medical histories, and public health data across the country to establish a unified platform for sharing anonymized medical data. Based on this, we will establish a data sharing incentive mechanism to encourage medical and research institutions to proactively upload anonymized data to the platform, while implementing a full process tracking mechanism for data usage to ensure that “privacy remains protected and data does not lie dormant.” At the same time, we will foster data sharing across regions and institutions, break down data silos, and create a unified national health data marketplace to drive innovative iterations of AI technologies in drug discovery, precision diagnostics, and other areas.
This reflection on data governance and sharing platforms resonated within the industry. Cao Peng similarly expressed that the National Health Commission should take the lead in building a national medical and health data sharing platform, leveraging privacy computing and blockchain technology to facilitate the secure sharing of medical data in a “usable but invisible” way. This eliminates data silos between healthcare organizations and provides high-quality, compliant data support for training and optimizing healthcare AI models.
What technical and institutional boundaries are needed for AI in healthcare?
Despite the bright prospects for AI in healthcare, potential risks and positioning issues remain the focus of industry concerns.
Zhang Wenhong, a member of the National Committee of the Chinese People’s Political Consultative Conference and director of the Department of Infectious Diseases at Fudan University Huashan Hospital, said in two sessions this year that patients using artificial intelligence for self-diagnosis may receive some health guidance, but there are also significant risks.
“AI can act as a ‘super assistant’ for doctors, but it cannot replace doctors,” said Wenhong Zhang.
In his view, ordinary patients lack medical training and blindly following diagnostic advice provided by AI systems could lead to misdiagnoses and delays in treatment.
Therefore, the appropriate role of artificial intelligence in the medical system should be as an “amplifier” of the capabilities of doctors. For example, AI can handle early disease screening, risk alerts, and data analysis, but the final diagnostic decision still needs to be made by a doctor.
While discussing the limitations of the technology, we should not overlook the challenges facing AI in healthcare in terms of data security and algorithmic transparency. Wang Jian’an, member of the National Committee of the Chinese People’s Political Consultative Conference, academician of the Chinese Academy of Sciences, and director of the Second Affiliated Hospital of Zhejiang University School of Medicine, stressed that the medical field is highly specialized and high-risk, and the application of artificial intelligence must prioritize safety.
“Medical care is related to human life, and the use of AI in medical care must prioritize safety,” said Wang Jian’an.
To turn this “baseline” into an enforceable industry standard, Liu Qingfeng suggested to journalists that an access and regulatory system for AI applications in basic public health needs to be established. Standards should be developed for AI-powered family physician and medical assistant products, with clearly defined requirements for data security and algorithmic accuracy. They should also establish quality assessment systems to continuously monitor application results and rigorously enforce privacy protection and data security requirements, with guidelines specifying scenarios, processes, and responsibilities.
Echoing Liu Qingfeng’s suggestion on access control, Dai Lizhong emphasized the need to quickly develop industry standards to define technical access standards, quality evaluation standards, and clinical application standards for AI medical products, with a view to improving support policies. Innovative regulatory models need to be developed to clarify the legal validity and division of responsibility for AI diagnostics and to establish a comprehensive “pre-access, in-process monitoring, and post-event accountability” regulatory system using AI technologies that enables dynamic monitoring of AI medical product applications. Beyond institutional constraints and technical oversight, human factors remain critical to ensuring a secure closed loop. Cao Peng proposed incorporating the ability to use medical AI into the competency evaluation system for doctors, implementing standardized training on AI applications for doctors, and promoting the spread of human-AI collaborative diagnosis and treatment models.
