Implementation Opinions of NMPA on "Artificial Intelligence + Drug Regulation"

GYJZ [2026] No. 6

To medical products administrations of all provinces, autonomous regions, and municipalities directly under the central government, and Xinjiang Production and Construction Corps, as well as all departments and affiliated institutions of the NMPA:

Since the implementation of the Drug Smart Regulation Action Plan, drug regulatory authorities at all levels have actively explored the use of information technology to enhance drug regulation capabilities, and have initially established a nationally-integrated drug smart-regulation system. Currently, the rapid development and iterative advancement of new-generation information technologies, such as AI, provide new tools and inject new impetus into smart regulation. To implement the Opinions of the State Council on In-depth Implementation of the "Artificial Intelligence +" Initiative and the Opinions of the General Office of the State Council on Comprehensively Deepening the Reform of Regulation of Drugs and Medical Devices to Promote the High-Quality Development of the Pharmaceutical Industry, seize the major strategic opportunity presented by the development of AI, promote the in-depth integration of AI with drug regulation, and accelerate the modernization of drug regulation, the following opinions are hereby formulated as follows.

I. General Requirements

Guided by Xi Jinping Thought on Socialism with Chinese Characteristics for a New Era, we will thoroughly implement the spirit of the 20th National Congress of the Communist Party of China and the successive plenary sessions of the 20th Central Committee, and earnestly follow General Secretary Xi Jinping's important instructions on drug regulation. We will adhere to information technology-led modernization of drug regulation, stay problem-oriented and systematic in thinking, balance development and safety, leverage the Smart Regulatory Information Platform as the main hub, strengthen system collaboration and open sharing, take data elements as the driver and scenario applications as the traction. We will advance the innovative application of AI in the whole lifecycle of drug regulation, improving the level of "one-stop online services, one-network unified governance, and one-network collaboration" through automation, precision, collaboration, and intelligence. We will build a high-level, integrated national drug smart regulatory system to provide strong data-driven support for comprehensively deepening drug regulation reform.

By 2030, an integrated innovation system for AI + drug regulation will be preliminarily established. The operational management mechanism of "AI + Drug Regulation" will be essentially formed, the computing power support infrastructure will become more integrated and efficient, and high-quality datasets, vertical large models, and intelligent agents that meet regulatory intelligence needs will be developed. AI will be effectively applied in various scenarios such as review and approval, supervision and inspection, testing and surveillance, and government services. The efficiency of human-machine collaboration will be significantly enhanced, and digital intelligence-driven regulatory capabilities throughout the whole lifecycle will reach new heights. By 2035, a new pattern of smart drug safety governance featuring digital-driven, intelligent and agile, independent controllability and ecological collaboration will be basically formed.

II. Focus on Key Scenarios of Smart Regulation and Empower the Implementation of Regulatory Reforms through Digital-Intelligence

(I) Build a human-machine collaborative intelligent review and approval system. Promote the standardization and structuring of electronic submission of application dossiers, improve the review and approval knowledge base, and accelerate the R&D and application of large models and intelligent agents for the review of "drugs, medical devices, and cosmetics." This will efficiently empower scenarios such as intelligent product classification, task assignment, documents review, knowledge retrieval, problem identification, report generation, and certificate delivery, thereby significantly improving the quality and efficiency of review and approval. Focusing on high-frequency scenarios in the review and approval work of local regulatory authorities, and following the principle of NMPA guidance and the collaborative division of responsibilities among provincial medical products administrations (PMPAs), we will accelerate the implementation of intelligent applications in key scenarios, using pilot projects to drive broader adoption. These scenarios include the review of Class II medical devices, post-market changes for drugs, the filing of general cosmetics, and approval of manufacturing and distributing licenses. This will strengthen transition and sharing of achievements while avoiding low-level, redundant development. Further refine the working mechanism for AI-assisted review and approval work. With ensuring product safety and efficacy as the baseline and improving review quality and efficiency as the focus, we will establish and refine a human-machine collaboration mechanism featuring "data-intelligence empowerment, manual review, and full-process traceability," and accelerate the development of an efficient, safe, and manageable intelligent review and approval system.

(II) Enhancing intelligent regulatory capabilities across the entire chain. In the research and development stage, continue to advance the standardization of clinical trial data governance. Research and formulate technical guidelines for electronic records of clinical trials, computerized system validation, and other supporting regulations, improve the technical guidelines system, and leverage clinical trial big data to enhance regulatory effectiveness. In the production stage, we will continuously refine the digital intelligence regulatory mechanism for high-risk products such as vaccines, blood products, and special drugs. We will improve the regulatory model that combines on-site inspections with off-site surveillance, develop and deploy intelligent risk monitoring agents, and dynamically monitor quality and safety risks in the production process based on real-time analysis of enterprise production process data, such as surveillance video, images, and IoT sensor data. In the distribution and use stage, we will promote the digital and intelligent upgrading of the drug traceability system, further urge enterprises to fulfill primary traceability responsibilities, and enable platform enterprises to enhance their technical support capabilities and service levels. We will accelerate the serialization and coding of all marketed products to achieve full-process traceability across production, distribution, and use. Relying on the traceability collaboration platform, accelerate the filing of traceability coding rules for all products and build a multi-code relational mapping database linking drug traceability codes with product barcodes, medical insurance codes, and other codes. Strengthen information-based traceability regulation throughout the entire process for key products, and deepen trigger-based traceability regulation. Develop technical guidelines for typical applications of Unique Device Identification (UDI) for medical devices in the production, distribution, and use stages, and actively explore its application across the entire-chain regulation.

(III) Promoting the digital intelligence upgrade of the risk regulatory system. Promote the multi-source aggregation, intelligent analysis, graded dissemination, and traceable tracking of risk clues. Advance data-driven regulation, and refine the risk consultation mechanism featuring "monitoring and early warning — consultation and analysis — directive handling — tracking and retrospective review". Strengthen capabilities in risk perception, intelligent early warning, and coordinated response for key products, enterprises, and stages, thereby comprehensively enhancing the effectiveness of risk regulation. Further advance the Smart Drug Testing Initiative, promote the development of an integrated, digital and intelligent testing system, improve testing efficiency, accuracy, and the ability to identify risk signals, and encourage exploration of robotics technology in testing. Upgrade and refine the monitoring and evaluation system for "drugs, medical devices, and cosmetics," and promote the intelligent transformation of scenarios such as report submission, review and evaluation, intelligent analysis, risk warning, and cross-level collaboration, thereby enhancing the level of intelligent monitoring and evaluation. Establish an intelligent analysis and early warning system for complaints and reports. Coordinate the upgrading of the safety risk and public opinion monitoring system for the online sales of "drugs, medical devices, and cosmetics," and refine the working mechanism to achieve full-domain real-time monitoring, accurate assessment, scientific early warning, and effective response. Strengthen the intelligent analysis of traceability data, build traceability risk screening and early warning models, and improve the level of intelligent monitoring for distribution risks. Enhance the integration, governance, and analysis of big data for whole lifecycle regulation. Focus on high-risk products and key scenarios, develop intelligent risk regulation models for quality safety, distribution anomalies, and online sales monitoring, create intelligent risk monitoring agents, and develop dynamic risk profiles for key products, key enterprises, and key stages.

(IV) Advancing the intelligent and standardized approach to inspections and law enforcement. Deepen the Smart Inspection Initiative, integrate and upgrade the inspection system, and build an integrated, intelligent, comprehensive management platform for smart inspections. Based on the big data of product and credit archives for "drugs, medical devices, and cosmetics," conduct risk assessment and determine inspection targets, frequencies, and plans scientifically and rationally according to risk levels, reducing repetitive inspections and implementing precise inspections. Encourage provincial-level drug regulatory departments to build unified systems for supervision, inspection, and law enforcement case handling, and strengthen digital intelligence support for city and county-level regulatory personnel in the supervision, inspection, and enforcement related to "drugs, medical devices, and cosmetics." Deepen the application of AI to support real-time information queries on regulated entities, real-time data capture during regulatory processes, intelligent discovery of violation clues, and automatic generation of documents and reports. Accelerate the standardization of inspection and law enforcement workflows, and improve the effectiveness and consistency of on-site inspection and law enforcement. Strengthen mobile inspection and law enforcement capabilities, implement "QR code-based entry into enterprises", and realize "fingertip management and palm-top inspection".

(V) Enhancing collaborative regulatory effectiveness. Enhance cross-region, cross-level, and cross-department collaborative regulatory capabilities using digital intelligence technology, focusing on solving prominent issues such as inadequate collaboration mechanisms, inefficient business flows, lack of information sharing, and difficulties in closing the loop on issue resolution. Leveraging the Smart Regulatory Information Platform, we will build an efficient, intelligent, multi-stakeholder national integrated business collaboration system, improve the management mechanism for collaborative item lists, standardize business collaboration processes, rules, and interface standards, and focus on key areas and critical stages such as clinical trials, registration verification, cross-provincial entrusted manufacturing, and the supervision of products selected in centralized volume-based procurement. We will advance intelligent assignment, full traceability, and closed-loop management of cross-level and cross-region collaborative operations. Strengthen cross-departmental regulatory information sharing and business coordination to promote the coordinated development and governance of "three medical linkages" (medical care, medical insurance, and pharmaceuticals), effectively supporting joint inspections, case handling, administrative-criminal law enforcement coordination, and clue disposal, and enhance collaborative regulatory effectiveness.

(VI) Enhancing the intelligence level of government services. Implement the ongoing requirements for the "Efficiently Handling One Matter (integrated government services)" initiative and strengthen departmental collaboration and service integration. Accelerate the development of "AI + Government Services," improve the policy service knowledge base, and integrate data such as policies and regulations, service guidelines, frequently asked questions, online consultations, user feedback, and historical processing records. Refine policy requirements, policy tags, and push conditions, and optimize algorithm models. Provide services such as intelligent Q&A, intelligent guidance, intelligent form pre-filling, and intelligent assisted handling for enterprises and the public, advancing the intelligence, precision, and convenience of government services.

(VII) Promoting collaborative digital intelligence development between regulation and industry. Focusing on the requirements for intelligent regulation, encourage and guide the industry to accelerate its digital intelligence transformation and upgrade, enhancing the digital intelligence level across the entire process, including drug research and development, manufacturing, quality testing, and post-market surveillance and evaluation. Accelerate the development of guiding principles for the standardized application of AI in the pharmaceutical industry to meet the needs of emerging technologies in the industry. Advance the full digital intelligence transformation of manufacturing and testing processes for high-risk products such as blood products and traditional Chinese medicine injections. Develop supporting regulatory requirements and gradually expand this to other products, guiding the industry to improve the quality control capabilities across the entire process in accordance with regulatory norms.

III. Seizing the New Trends in AI Development and Strengthening the "AI + Drug Regulation" Foundational Support

(I) Promoting the development of high-quality drug regulation datasets. Adhere to the principle of "scenario-driven, urgent needs prioritized" and, focusing on the core business scenarios of the entire drug lifecycle regulation and the practical needs of AI applications, proceed with the phased and stepwise development of high-quality drug regulation datasets. Further improve the nationwide integrated drug regulation data resource system, with national and provincial-level data centers as hubs. We will improve the data aggregation and governance system based on product archives, enterprise credit records, legal and regulatory libraries, and typical case databases, improve the data aggregation and governance system. This will enhance the accuracy, consistency, and usability of the data, providing foundational support for the development of high-quality datasets. Focus on the training, fine-tuning, and practical application of vertical large models for drug regulation. Clearly define data formats, quality, and content requirements according to specific scenarios, and develop scientific and unified collection standards and labeling guidelines. Conduct multi-source data fusion governance, professional annotation, and knowledge extraction. Build both general and specialized knowledge bases for drug regulation, forming a layered, classified, dynamically updated, and fully traceable across the entire lifecycle. Under the strict condition of ensuring security and privacy, orderly promote the compliant and efficient application of knowledge bases and high-quality datasets in scenarios such as model training, knowledge inference, and decision support.

(II) Strengthening the AI application support system. Adhere to business-driven approach and comprehensively advance the training, deployment, and application of large models in the field of drug regulation. Leverage existing smart regulation infrastructure to build a large model application and algorithm management platform. Develop model application guidelines and security standards, promote the co-construction and sharing of common technical components, enhance model and algorithm management capabilities, and foster technological interoperability, resource sharing, and ecosystem collaboration. Promote the deep integration of AI with business information systems, accelerating the large-scale implementation of AI-assisted regulatory scenarios. Focusing on the entire drug lifecycle regulation, construct a multi-agent collaborative mechanism, improve system linkage and business collaboration frameworks, and drive the intelligent upgrade of drug regulation capabilities.

(III) Strengthening computing power infrastructure. The NMPA will coordinate the planning of a multi-level intelligent computing power resource coordination system, with national and provincial-level regulatory departments advancing the provision of smart computing resources as needed. Create a standardized and scalable intelligent computing power foundation to meet the smart application needs of various network domains, such as the Internet, government extranet, and government intranet. Enhance cross-domain collaboration and disaster recovery capabilities, gradually forming a "co-building, co-governance, and co-sharing" deployment pattern to improve computing power support capabilities, ensuring sustained and stable support for regulatory intelligence.

(IV) Fortifying the security protection system. Strictly implement the security responsibility system, upgrade the cybersecurity protection system, and improve mechanisms for cybersecurity situational awareness, information sharing, joint analysis, threat warning, and traceability and attribution. Utilize AI technologies to enhance proactive cybersecurity defense capabilities and build an intelligent, collaborative protection system. Establish and improve the data security management system, clarify core and important data catalogs, and enhance the technical framework for data security protection. Strengthen AI risk monitoring and evaluation, establish algorithm transparency requirements and model validation standards, and strengthen security capabilities for models, algorithms, data resources, infrastructure, and application systems. Enhance AI application risk assessments, response, prevent confidential and sensitive information from entering non-confidential models, and promote the compliant, transparent, and trustworthy use of AI applications.

(V) Improving the construction and operation management mechanism. Adhere to the supportive role of AI in the drug regulation field, clearly define the functional boundaries and responsibilities for large models and various intelligent support applications, and avoid unapproved deployments, fragmented construction, and overlapping efforts. Establish a dedicated mechanism to oversee the governance of AI applications in drug regulation, coordinating model construction access, safety reviews, and scenario compliance reviews and more. Develop a management system for "AI + Drug Regulation," clarifying responsibilities and work standards. Improve the model and algorithm filing management system, establish basic principles and technical standards, and conduct validation and evaluation of the effectiveness and reliability of models and their supporting applications. Strengthen the management of training data, fine-tuning data, and knowledge bases, ensuring the legality of their sources, accuracy of content, compliance with use, and traceability throughout the process. Explore the authorized operation of public drug regulation data, build dedicated public data areas, promote field-specific and scenario-based authorization, and enhance the development and utilization of public drug regulation data.

IV. Organizational Implementation

Drug regulatory authorities at all levels shall deeply understand the new trends in AI development and regard it as a key lever to support the comprehensive deepening of drug regulation reform and as a strong support for enhancing drug regulation capabilities. These authorities shall coordinate and align relevant plans, increase investments, and promote the application of AI in frontline regulation. The approach shall be "promoting construction through application and integrating development with application," ensuring that AI plays a practical role in regulation. There shall be a focus on demonstration and leadership, targeting the challenges and bottlenecks in regulatory business, and deepening the innovative applications of smart regulation to effectively empower business innovation. Strengthen regulatory science research to provide technological support for "AI + Drug Regulation," and promote the translation and application of relevant major scientific and technological projects. Increase training efforts to enhance the digital thinking, digital skills, and digital literacy of the regulatory workforce.

NMPA

March 11, 2026

(April 2, 2026)

国家药监局关于“人工智能+药品监管”的实施意见

国药监综〔2026〕6号

各省、自治区、直辖市和新疆生产建设兵团药品监督管理局，局机关各司局、各直属单位：

药品智慧监管行动计划实施以来，各级药品监管部门积极探索利用信息化手段提升药品监管能力，初步构建了全国一体化药品智慧监管体系。当前，人工智能等新一代信息技术的快速发展与迭代演进，为智慧监管提供新手段、注入新动能。为贯彻落实《国务院关于深入实施“人工智能+”行动的意见》《国务院办公厅关于全面深化药品医疗器械监管改革促进医药产业高质量发展的意见》，抢抓人工智能发展重大战略机遇，推进人工智能与药品监管深度融合发展，加快推动药品监管现代化，提出以下意见。

一、总体要求

以习近平新时代中国特色社会主义思想为指导，深入贯彻党的二十大和二十届历次全会精神，全面贯彻习近平总书记关于药品监管的重要指示批示精神，坚持以信息化引领药品监管现代化，坚持问题导向、系统思维，统筹发展和安全，发挥智慧监管平台总枢纽作用，强化系统协同和开放共享，以数据要素为驱动、以场景应用为牵引，深入推进人工智能在药品全生命周期监管中的创新应用，通过自动化、精准化、协同化、智能化提升“一网通办、一网统管、一网协同”水平，打造高水平全国一体化药品智慧监管体系，为全面深化药品监管改革提供有力数智支撑。

到2030年，初步构建药品监管与人工智能融合创新体系，“人工智能+药品监管”运行管理机制基本形成，算力支撑底座更加集约高效，形成满足监管智能化需要的高质量数据集、垂直大模型和智能体，人工智能在审评审批、监督检查、检验监测、政务服务等场景中有效应用，人机协同效率显著提升，全生命周期数智化监管能力迈上新台阶。到2035年，基本形成数智驱动、智能敏捷、自主可控、生态协同的智慧化药品安全治理新格局。

二、紧扣智慧监管重点场景，数智赋能监管改革落地生效

（一）构建人机协同智能审评审批体系。推动申报资料电子提交标准化、结构化，完善审评审批知识库，加快“两品一械”审评审批大模型与智能体研发应用，高效赋能产品智能分类、任务分配、资料审查、知识检索、问题识别、报告生成、制证送达等场景，显著提升审评审批质效。聚焦地方监管部门在审评审批工作中的高频场景，按照国家局指导，省级药品监管部门分工协作的原则，以点带面，加快推进第二类医疗器械审评、药品上市后变更备案、普通化妆品备案、生产经营许可审批等重点场景智能化应用落地，强化成果转化与共享，避免低水平重复建设。进一步完善人工智能辅助审评审批工作制度，以保障产品安全有效为底线、以提升审评审批质效为重点，建立健全“数智赋能、人工复核、全程留痕”的人机协同机制，加快建设高效、安全、可控的智能化审评审批体系。

（二）提升全链条智能化监管能力。在研制环节，持续推进临床试验数据治理规范化，研究制定临床试验电子化记录技术指南、计算机化系统验证指南等配套规范，完善技术指南体系，利用临床试验大数据提升监管效能。在生产环节，持续完善疫苗、血液制品、特殊药品等高风险品种生产数智化监管机制，完善现场检查与非现场监管相结合的监管模式，研发部署风险监控智能体，基于企业对生产过程监控视频、图像、物联感知等数据的实时分析结果，动态监测生产过程质量安全风险。在流通使用环节，推动药品追溯体系数智化升级，进一步督促企业落实追溯主体责任，平台企业提升技术支撑能力与服务水平，加快推进全部在产品种赋码，实现生产、流通、使用全过程可追溯。依托追溯协同平台加快推进全品种追溯编码规则备案，构建药品追溯码与商品条码、医保编码等多码关联映射数据库。加强重点品种全过程信息化追溯监管，深化触发式追溯监管。制定医疗器械唯一标识在生产经营使用环节典型应用技术指南，积极探索医疗器械唯一标识在全链条监管中的应用。

（三）推动风险监管体系数智升级。推动风险线索多源归集、智能研判、分级下达和留痕跟踪，推进数据驱动监管，完善“监测预警—会商研判—指令处置—跟踪回溯”的风险会商机制，强化对重点品种、重点企业和重点环节的风险感知、智能预警和协同处置能力，全面提升风险监管效能。深入开展智慧药检建设行动，推进一体化、数智化检验体系建设，提升检验效率、准确性和风险信号识别能力，鼓励探索机器人技术在检验中的应用。升级完善“两品一械”监测评价系统，推动报告上报、审核评价、智能分析、风险预警、跨层级协作等场景的数字化、智能化，提升智能化监测评价水平。建设投诉举报智能分析预警系统。统筹推进“两品一械”网络销售安全风险监测和舆情监测体系升级，完善工作机制，实现全域实时监测、准确研判、科学预警和有效处置。强化追溯数据智能分析，构建追溯风险筛查与预警模型，提升流通风险智能监测水平。加强全生命周期监管大数据的整合治理与分析挖掘，聚焦高风险品种和重点场景，研发质量安全、流通异常、网售监测等智能风险监管模型，开发风险监测评估智能体，绘制重点品种、重点企业、重点环节的动态风险画像。

（四）推进检查执法智能化规范化。深入开展智慧检查建设行动，整合升级检查系统，构建集约智能的智慧检查综合管理平台。基于“两品一械”品种档案与信用档案大数据开展风险研判，依据风险等级科学合理确定检查对象、检查频次和检查方案，减少重复检查、实施精准检查。鼓励省级药监部门建设统一的监督检查与执法办案系统，加强对市县监管人员“两品一械”监督检查和稽查执法的数智化支撑。加强人工智能深化应用，支撑监管对象信息实时查询、监管过程信息实时录入、问题线索智能发现、文书和报告自动生成等，加快推进检查执法业务流程标准化，提升现场检查执法效能和规范性。加强移动检查执法建设，推行“扫码入企”，实现“指尖管、掌上查”。

（五）提升协同监管效能。以数智技术增强跨区域、跨层级、跨部门协同监管能力，着力破解协同机制不健全、业务流转低效、信息不共享、处置难闭环等突出问题。依托智慧监管平台建设高效智能、多方联动的全国一体化业务协同系统，健全协同事项清单管理机制，规范业务协同流程规则和接口标准，聚焦临床试验、注册核查、跨省委托生产、集采中选产品监管等重点领域和关键环节，推进跨层级、跨区域协同业务的智能分派、全程可溯与闭环管理。加强跨部门监管信息共享和业务联动，促进“三医”协同发展与治理，有力支撑联合检查、案件查办、行刑衔接、线索处置等业务场景，提升协同监管效能。

（六）提升政务服务智能化水平。落实常态化推进“高效办成一件事”要求，加强部门协同和服务集成。加快推进“人工智能+政务服务”建设，完善政策服务知识库，整合政策法规、办事指南、常见问题、网上咨询、用户评价和历史办理记录等数据，细化政策要求、政策标签、推送条件，优化算法模型，为企业和群众提供智能问答、智能引导、智能预填、智能帮办等服务，推进政务服务智能化、精准化、便利化。

（七）促进监管与产业数智化协同发展。围绕智能化监管要求，鼓励引导产业加快推动数智化转型升级，提升药品研制、生产制造、质量检验、上市后监测评价等全过程数智化水平。加快研究制定人工智能在医药产业规范应用的指导原则，适应产业新技术应用发展需要。推进血液制品、中药注射剂等高风险品种生产、检验全程数智化，研究制定配套监管要求，逐步拓展到其他品种，引导产业按规范提升全过程质量管控能力。

三、把握人工智能发展新趋势，筑牢“人工智能+药品监管”基础支撑

（一）推进药品监管高质量数据集建设。坚持“场景驱动、急用先行”原则，围绕药品全生命周期监管核心业务场景及人工智能应用的实际需求，分阶段分步骤推进药品监管高质量数据集建设。进一步完善全国一体化药品监管数据资源体系，以国家、省两级数据中心为枢纽，以品种档案、企业信用档案、法律法规库、典型案例库等为基础，健全数据汇聚与治理体系，提升数据准确性、一致性与可用性，为高质量数据集建设提供基础支撑。聚焦药品监管垂直大模型的训练、微调与落地应用，依场景明确数据格式、质量、内容要求，制定科学统一的采集规范与标注指南，开展多源数据融合治理、专业标注与知识抽取，构建药品监管领域通识与专识知识库，形成分层分类、动态更新、全生命周期可追溯的高质量数据集。在严格保障安全与隐私前提下，有序推动知识库和高质量数据集在模型训练、知识推理、辅助决策等场景的合规高效应用。

（二）强化人工智能应用支撑体系。坚持业务引领，统筹推进药品监管领域大模型的训练、部署和应用。依托现有智慧监管基础设施，建设大模型应用及算法管理平台，制定模型应用指引与安全规范，推动共性技术组件的共建共享，提升模型与算法管理能力，促进技术互通、资源共享与生态协同。推动人工智能与业务信息系统深度融合，加速人工智能辅助监管场景规模化落地。围绕药品全生命周期监管，构建多智能体协同机制，完善系统联动与业务协同体系，推动药品监管能力智能化升级。

（三）加强算力基础设施建设。国家局统筹规划多级智能算力资源协同体系，国家、省两级监管部门按需推进智算资源供给。打造标准化、可扩展的智能算力底座，满足互联网、政务外网、政务内网等不同网络域的智能应用需求。提升跨域协同与容灾能力，逐步形成“共建、共治、共享”的部署格局，提升算力支撑能力，为监管智能化提供持续、稳定保障。

（四）筑牢安全防护体系。严格落实安全责任制，升级网络安全防护体系，健全网络安全态势感知、信息共享、联合研判、威胁预警、追踪溯源机制，运用人工智能技术提升网络安全主动防护能力，构建智能化、协同化防护体系。建立健全数据安全管理体系，明确核心和重要数据目录，完善数据安全保护技术体系。强化人工智能风险监测评估，制定算法透明度要求和模型验证规范，加强模型算法、数据资源、基础设施、应用系统等安全能力建设，强化人工智能应用风险评估和监测处置，防止涉密信息和敏感信息输入非涉密模型，推动人工智能应用合规、透明、可信赖。

（五）完善建设运行管理机制。坚持人工智能在药品监管领域的辅助型定位，明确大模型及各类智能辅助应用的功能边界和责任主体，避免未经审查即部署、多头建设、重复建设等行为。成立专设机制负责药品监管人工智能应用治理工作，统筹模型建设准入、安全审查、场景合规性审查等，制定“人工智能+药品监管”管理制度，明确职责分工和工作规范。完善模型和算法备案管理制度，制定基本准则与技术规范，对模型及其辅助应用开展有效性、可靠性验证与评估。加强训练数据、微调数据和知识库等数据资源管理，确保来源合法、内容准确、使用合规、全程可追溯。探索药品监管公共数据授权运营，建设公共数据专区，推动分领域、依场景授权，加强药品监管公共数据开发利用。

四、组织实施

各级药品监管部门要深刻认识人工智能发展新趋势，将其作为支撑全面深化药品监管改革的重要抓手和提升药品监管能力的有力支撑，统筹衔接好相关规划，加大投入，推动人工智能在一线监管中的应用，以用促建，建用结合，真正发挥人工智能在监管中的实效。要强化示范引领，聚焦监管业务的难点、堵点，深化智慧监管创新应用，有效赋能业务创新。加强监管科学研究对“人工智能+药品监管”的科技支撑，推动相关重大科技项目落地与转化应用。加大培训力度，提升干部队伍数字思维、数字技能和数字素养。

国家药监局

2026年3月11日

（2026-04-02）

如英文译本与中文解释上遇有分歧时，以中文文本为准。