1 National Science Library (Chengdu), Chinese Academy of Sciences, Chengdu 610299, China
2 Department of Information Resources Management, School of Economics and Management, University of the Chinese Academy of Sciences, Beijing 100191, China
[Objective/Significance] This study aims to conduct an in-depth analysis of the current research status of generative AI (GenAI) governance domestically and internationally, identify research hotspots, and explore the research landscape. [Method/Process] Based on literature related to GenAI security governance from CNKI and WoS, this paper employs a combination of scientometric analysis and BERTopic topic modeling to reveal the current state of GenAI security governance research from two major perspectives: institutional characteristics and thematic content analysis, with the goal of providing recommendations for governance practices in China. [Results/Conclusions] The findings indicate that partial collaborative innovation networks have formed domestically and internationally. Overall, risk analysis is conducted across five governance scenarios: data security, online public discourse ecosystems, academic ethics, educational settings, and healthcare. Governance solutions derived include strengthening ideological guidance and technical education, establishing sound laws and regulations, improving regulatory safeguard systems, enhancing technical monitoring capabilities, and forming agile governance models.
Chen Lu, Chen Fang.
Quantitative Analysis of the Research Status on Generative AI (GenAI) Security Governance Empowered by Digital Intelligence[J].
Science Focus, 2026, 21(1): 35-55 DOI:10.15978/j.cnki.1673-5668.20250709
本文以中国知网(CNKI)数据库中“北大核心”“CSSCI”“AMI”“CSCD”作为中文文献数据来源,以Web of Science(简称“WoS”)核心集数据库中的SCI-EXPANDED、SSCI、CPCI-S、CPCI-SSH、ESCI索引作为英文文献数据来源,检索以“GenAI治理”研究为主题的文献,基于国内外文献的角度,对GenAI治理的研究现状进行科学计量分析,检索所有截至2025年5月16日的文献,检索措施见表1。中文文献经人工判读过滤后最终保留206篇,英文文献去除与主题不符合的、非英文的以及在作者、期刊、发文机构等关键字段缺少信息的条目后最终保留83篇作为最终的分析数据。
在国外的核心发文机构中,宾夕法尼亚大学(University of Pennsylvania)与匹兹堡大学(University of Pittsburgh)大学发文量最高,其中宾夕法尼亚大学主要从社会影响的角度探讨GenAI在社会经济发展、教育等领域的治理研究[33-34],匹兹堡大学则重点关注GenAI在教育领域的负责任使用,以及在医疗环境中的伦理管理框架[5,33-34]。国外7所核心发文机构分别来自美国、澳大利亚、英国和新加坡,且都是世界著名的研究型大学。其中哈佛大学(Harvard University)、悉尼大学(University of Sydney)、新加坡国立大学(National University of Singapore)、宾夕法尼亚大学的法学专业与医学专业均在2025年QS世界大学学科排名前50,诺丁汉大学(University of Nottingham)的法学专业在2025年QS世界大学学科排名前100,哈佛大学、新加坡国立大学、宾夕法尼亚大学的商业与管理研究专业在2025年QS世界大学学科排名前50。据此可见,国外对GenAI治理的研究集中在法学、商学和医学专业背景强势的院校,结合上文中对高发文单位研究内容的简单概述得出,国内外对于GenAI治理的研究具有从法学视角出发的共性,相比于国内,国外的研究更聚焦,与研究背景结合得更紧密。
国外机构合著网络 (图4) 有154个合著节点,519条边,网络平均聚类系数为0.946,平均路径长度为1.634,平均度为6.74,根据式(2)的计算方式,图4的边密度的理论值p国外=2×519/154×153≈0.0441,且国外机构合著网络的群居系数为0.946,显著高于理论值,表明本网络具有较强的局部聚类特征。根据式 (3) 的计算方法,平均路径长度理论值L国外=ln154/ln6.74≈2.64,而实际值(1.634)远小于理论值,这种聚类系数高、平均路径小的网络特征说明国外机构合著网络也呈现出明显的小世界特征。其中纽约大学(New York University)、麻省大学(University of Massachusetts)、阿姆斯特丹自由大学(Vrije Universiteit Amsterdam)、密歇根州立大学(Michigan State University)、马普学会(Max Planck Society)、麻省理工学院(Massachusetts Institute of Technology,简称MIT)、 挪威经济学院(Norwegian School of Economics,简称NHH)、图卢兹经济学院(Toulouse School of Economics)、哥本哈根商学院(Copenhagen Business School)、微软(Microsoft)、剑桥大学(University of Cambridge)等(图4中蓝色部分)形成了紧密的合作团体。
由表4和表5可以看出,国内中介中心性前3的是北京大学、清华大学、华东政法大学,国外是纽约大学(New York University)、哈佛大学、蒙纳士大学(Monash University),说明这些单位在国内外GenAI治理研究领域中发挥着关键的连接作用,在合作机会与数据共享上具有优势。国内特征向量中心性前3的是北京大学、北京师范大学以及南开大学,国外特征向量中心性前3的是蒙纳士大学(Monash University)、昆士兰大学(University of Queensland)、纽约大学(New York University)。国内点度中心性较高的是北京大学、南京师范大学、北京师范大学,国外点度中心性较高的是蒙纳士大学(Monash University)、昆士兰大学(University of Queensland)、纽约大学,同时这6家单位的3项指标基本都位列国内外排名的前10,表明这些机构在领域内处于合作活跃度高、影响力扩散快,在国内外本领域的研究中处于绝对核心地位,资源整合能力强,研究通过直接合作快速传播,容易形成学术集群,引领研究趋势,具有一定的学术生态主导力。
研究主题的分析采用基于机器算法的BERTopic主题模型对中英文文献的摘要进行主题建模,在文本分词阶段,对中文文本采用了百度开发的LAC(lexical analysis of Chinese)中文分词工具[38]。该工具基于PaddlePaddle 深度学习平台,能够实现分词、词性标注与命名实体识别,具有较高的分词准确率,适合对科技文献进行深入分析[39]。英文文本采用了常用的Spacy分词工具。经过多轮实验,最终确定模型参数,结果见表6。在设定的参数条件下,识别出7个中文主题和5个英文主题,且都实现了较好的建模效果,中文主题一致性指数为0.39,英文主题一致性指数为0.73,各主题的文档数目以及占全部文档的比例见表7。
(RenL M. Mechanisms and practices of generative artificial intelligence empowers philosophical and social sciences research[J]. Observation and Ponderation, 2025(5): 15-25.)
(ChengQ, LiuH D. The construction and risk prevention of ideological and political education scene based on AGI[J]. Journal of National Academy of Education Administration, 2024(8): 87-95.)
(ShenY, FangS M, LiuC C, et al. The governance of generative artificial intelligence in educational applications: cases and reflections[J]. Open Education Research, 2024, 30(6): 39-47.)
[7]
Vani LakshimiR, ClareR S, KamathA. Demystifying the ethical framework for generative AI in healthcare: a data science perspective[C]// Artificial Intelligence in Healthcare: First International Conference, AIiH 2024, Swansea, UK, September 4-6, 2024, Proceedings, Part II. Berlin, Heidelberg: Springer-Verlag, 2024: 279-289.
[8]
JacksonB R, RashidiH H, LennerzJ K, et al. Ethical and regulatory perspectives on generative artificial intelligence in pathology[J]. Archives of Pathology & Laboratory Medicine, 2025, 149(2): 123-129.
(TangZ Y, XieY J. The risk of copyright infringement in the use of generative artificial intelligence data and its governance[J]. China Publishing Journal, 2024(21): 56-61.)
(YaoY F, HuF, XuZ Z. Opportunities, risks, and countermeasures for empowering government data security governance with generative artificial intelligence[J]. Think Tank: Theory & Practice, 2024, 9(4): 1-11.)
[15]
YeX B, YanY H, LiJ, et al. Privacy and personal data risk governance for generative artificial intelligence: a Chinese perspective[J]. Telecommunications Policy, 2024, 48(10): 102851.
(ChuJ W, DuX X. Research review of generative artificial intelligence empowering knowledge production in scientific research[J]. Journal of Academic Libraries, 2024, 42(3): 108-117.)
(PanL. A comprehensive review of ethical issues in generative artificial intelligence: a bibliometric and visualization analysis based on CiteSpace[J]. Journal of Kunming University of Science and Technology (Social Sciences), 2025, 25(2): 57-68.)
(ZhangX Y. A review of researches on generative artificial intelligence data: risks, challenges, and governance[J]. Library and Information Service, 2025, 69(9): 136-148.)
(ZhouZ F, QiuJ P, WeiK Y. From bibliometrics to “five-metrics”: the evolution and development of bibliometrics methods[J]. Journal of Intelligence, 2021, 40(10): 171-178.)
[24]
邱均平. 科学计量学[M]. 北京: 科学出版社, 2016.
[25]
(QiuJ P. Scientometrics[M]. Beijing: Science Press, 2016.)
(WassermanS, FaustK. Social network analysis: methods and applications[M]. Beijing: China Renmin University Press, 2025.)
[28]
GrootendorstM. BERTopic: neural topic modeling with a class-based TF-IDF procedure[EB/OL]. [2025-07-01].http://arxiv.org/abs/2203.05794.
[29]
DevlinJ, ChangM W, LeeK, et al. BERT: pre-training of deep bidirectional transformers for language understanding[EB/OL]. [2025-07-01].http://arxiv.org/abs/1810.04805.
[30]
AbuzayedA, Al-KhalifaH. BERT for arabic topic modeling: an experimental study on BERTopic technique[J]. Procedia Computer Science, 2021, 189: 191-194.
(LuJ W, ChenY. Research on international open science topic mining based on BERTopic model[J]. Journal of the National Library of China, 2025, 34(2): 99-113.)
(HuY L, YaoH L. Data governance and the“cage” for ChatGPT-like generative artificial intelligence: responsive research based on self-regulation by industry subjects[J]. Credit Reference, 2024, 42(2): 46-57.)
(TongY F. Internal management regulation of generative artificial intelligence technology risk[J]. Studies in Science of Science, 2024, 42(10): 2038-2046.)
(LiuX Q, DongW K. Dual governance of administrative law and criminal law: a new paradigm of generative artificial intelligence governance[J]. Exploration and Free Views, 2024(7): 127-136.)
(CongL X, LiY L. Work risks, copyright ownership and effective governance for artificial intelligence text-to-video large model[J]. Journal of Xinjiang Normal University (Philosophy and Social Sciences), 2024, 45(6): 101-111.)
(XiaoH J, ZhangL L. Theoretical deconstruction and governance innovation of ethical misconduct in large model[J]. Research on Financial and Economic Issues, 2024(5): 15-32.)
(HeY H, LiX. New challenges and countermeasures of false information governance in generative artificial intelligence: based on the agile governance perspective[J]. Governance Studies, 2024, 40(4): 142-156.)
[50]
邵红红. 生成式人工智能版权侵权治理研究[J]. 出版发行研究, 2023(6): 29-38.
[51]
(ShaoH H. Research on copyright infringement governance of generative artificial intelligence[J]. Publishing Research, 2023(6): 29-38.)
(ChenY D, WangW. From “priority for urgent needs” to “gradual improvement”: construction of generative artificial intelligence governance system[J]. E-Government, 2024(4): 113-124.)
(YuanY Q, ChenC F. Moralizing technology: human-machine value alignment and ethic in large language models[J]. Nanjing Journal of Social Sciences, 2024(6): 88-97.)
[58]
DotanR, ParkerL S, RadzilowiczJ G. Responsible adoption of generative AI in higher education: developing a “Points to Consider” approach based on faculty perspectives[EB/OL]. [2025-07-01]. http://arxiv.org/abs/2406.01930.
[59]
CapraroV, LentschA, AcemogluD, et al. The impact of generative artificial intelligence on socioeconomic inequalities and policy making[J]. PNAS Nexus, 2024, 3(6):191.
(YuD, LiZ F. Ethical issues and governance in generative artificial intelligence social experiment[J]. Studies in Science of Science, 2024, 42(1): 3-9.)
(ZengY L, GaoS, CaoG H. Characteristics and inspirations of interdisciplinary research collaboration networks for public health emergencies[J]. Information Science, 2025, 43(5): 147-159.)
[65]
JiaoZ, SunS, SunK. Chinese lexical analysis with deep Bi-GRU-CRF network[EB/OL]. [2025-05-27]. http://arxiv.org/abs/1807.01882.
(Comparative analysis of part-of-speech tagging between Jieba segmentation and LAC segmentation[EB/OL]. [2025-07-01]. https://developer.baidu.com/article/details/2968419.)
(YangS L, YuY H. Topic mining and evolution analysis of information resource management research based on BERTopic model[J]. Information Science, 2024, 42(8): 12-21.)
(ZhiZ F. Information content governance of large model of generative artificial intelligence[J]. Tribune of Political Science and Law, 2023, 41(4): 34-48.)
(ZhaoZ Y. Data security risks and countermeasures of generative artificial intelligence[J]. Information and Documentation Services, 2024, 45(2): 30-37.)
(TouX D. On the data security risk of generative AI and its responsive governance[J]. Oriental Law, 2023(5): 106-116.)
[76]
LiK G, WuH, DongY P. Copyright protection during the training stage of generative AI: industry-oriented U.S. law, rights-oriented EU law, and fair remuneration rights for generative AI training under the UN's international governance regime for AI[J]. Computer Law & Security Review, 2024, 55: 106056.
LiuH, LeiQ S. Data risk of generative artificial intelligence and its legal regulation[J]. Journal of Chongqing University of Posts and Telecommunications (Social Science Edition), 2024, 36(4): 40-51.)
[81]
YangF, AbedinM Z, QiaoY N, et al. Toward trustworthy governance of AI-generated content (AIGC): a blockchain-driven regulatory framework for secure digital ecosystems[J]. IEEE Transactions on Engineering Management, 2024, 71: 14945-14962.
(ChenR, JiangY H. Research on the governance of training data as the core driving force of generative artificial intelligence[J]. Information and Documentation Services, 2024, 45(4): 87-95.)
(ZhangX. Data risks and governance pathways of generative artificial intelligence[J]. Science of Law (Journal of Northwest University of Political Science and Law), 2023, 41(5): 42-54.)
(LiS. The governance path for data security in AIGC from the perspective of risk prevention: taking GPT class models as an example[J]. Journal of Xizang Minzu University (Philosophy and Social Sciences Edition), 2023, 44(6): 139-145.)
[88]
BelgodereB, DogninP, IvankayA, et al. Auditing and generating synthetic data with controllable trust trade-offs[J]. IEEE Journal on Emerging and Selected Topics in Circuits and Systems, 2024, 14(4): 773-788.
(AnX M, LongZ Q, KuangM M. Research on the theoretical framework and constituent element of large model data governance under the standardization perspective[J]. Information and Documentation Services, 2024, 45(6): 75-83.)
[93]
KarpouzisK. Plato’s shadows in the digital cave: controlling cultural bias in generative AI[J]. Electronics, 2024, 13(8): 1457.
(RanL, ZhangW. Deepfake information in AIGC: generation mechanisms and governance strategies: an analytical framework based on actor-network theory[J]. Journal of Information Resources Management, 2025, 15(2): 137-150.)
(DaiJ P, QinY Y. Ideological risks and countermeasures of generative artificial intelligence such as ChatGPT[J]. Journal of Chongqing University (Social Sciences Edition), 2023, 29(5): 101-110.)
(ShuaiJ Q, HanY. An analysis of the risk patterns, generating causes, and governance logic of ldeology in generative artificial intelligence[J]. Journal of Southwest Minzu University (Humanities and Social Sciences Edition), 2024, 45(3): 186-193.)
[102]
WesterlundM. The emergence of deepfake technology: a review[J]. Technology Innovation Management Review, 2019, 9(11): 39-52.
[103]
BroinowskiA, MartinF R. Beyond the deepfake problem: benefits, risks and regulation of generative AI screen technologies[J]. Media International Australia, 2024: 1329878X241288034.
(XuM, XiaoS Y. Generative technology and audiovisual manipulation: internal mechanisms and governance strategies for deepfake in age of AIGC[J]. Journal of Beijing University of Posts and Telecommunications (Social Sciences Edition), 2024, 26(4): 27-35.)
(Provisions on the administration of deep synthesis of internet information services[EB/OL].[2025-05-29].https://www.moj.gov.cn/pub/sfbgw/flfggz/flfggzbmgz/202307/t20230705_482071.html.)
(XieX F, QuC. Value alignment: research on the risks of AI cultural technology ethics and good governance in the AIGC era[J]. Journal of Lanzhou University (Social Sciences), 2024, 52(3): 147-156.)
(OuyangL J, ZhangY H. The propositional origin, generative mechanism and governance approach of ideological risk of generative AI application[J]. Academic Exploration, 2023(11): 7-16.)
[112]
JonesN. How to stop AI deepfakes from sinking society: and science[J]. Nature, 2023, 621(7980): 676-679.
[113]
Blanco-GonzálezA, CabezónA, Seco-GonzálezA, et al. The role of AI in drug discovery: challenges, opportunities, and strategies[J]. Pharmaceuticals, 2023, 16(6): 891.
(YaoJ Z. On ethical risks of generative artificial intelligence involvement in academic publishing and its regulation[J]. Publishing Research, 2024(11): 89-96.)
[116]
HutsonM. Robo-writers: the rise and risks of language-generating AI[J]. Nature, 2021, 591(7848): 22-25.
(ChenY, YangL Y, FengR. From content production to order reshaping: research on content production risk and regulation of generative AI publishing[J]. View on Publishing, 2024(22): 61-67.)
(ZouT L, XiongQ J. Risks of academic misconduct induced by generative AI and its governance[J]. Journal of University of Jinan (Social Science Edition), 2024, 34(6): 163-167.)
(WuD, WuH J. Potential risks of the application of general large model education and their avoidance: from the perspective of technology ethics[J]. Journal of East China Normal University (Educational Sciences), 2024, 42(8): 64-75.)
(ZhaoL L, YanZ M. Ecological ethical and risk reduction of generative artificial intelligence education applications[J]. Journal of Guizhou Normal University (Social Sciences), 2023(5): 151-160. )
(ZhangH B, XuL. Ethical risks and governance approaches of generative artificial intelligence in education: based on the practical investigation of the Russell University Group[J]. Modern Educational Technology, 2024, 34(6): 25-34.)
(TianX P, XiaoZ Q. Academic ethics and risk management of graduate research writing empowered by generative AI[J]. Modern Educational Technology, 2024, 34(8): 23-32.)
[131]
DengX N, JoshiK D. Promoting ethical use of generative AI in education[J]. ACM SIGMIS Database: the DATABASE for Advances in Information Systems, 2024, 55(3): 6-11.
(WangY M, WangD, WangH J, et al. Research on risk governance of AIGC educational application ethics based on risk regulation[J]. China Educational Technology, 2023(11): 83-90.)
(ZhouH Y, ChangS L. Generative AI embedded in higher education: future scenarios, potential risks and solutions[J]. Modern Education Management, 2023(11): 1-12.)
[136]
WangY Y, LiuC, ZhouK Y, et al. Towards regulatory generative AI in ophthalmology healthcare: a security and privacy perspective[J]. The British Journal of Ophthalmology, 2024, 108(10): 1349-1353.
(XuL Y. Research on security risk and governance path of large models[J]. Journal of Information Security Research, 2024, 10(10): 975-980.)
[139]
ChenY, EsmaeilzadehP. Generative AI in medical practice: in-depth exploration of privacy and security challenges[J]. Journal of Medical Internet Research, 2024, 26: e53008.
[140]
IliuțăM E, TrentesauxD, PopE, et al. Integration of generative artificial intelligence in medicine:first steps toward an ethical risk management methodology[C/OL]// 2024 E-Health and Bioengineering Conference (EHB). IASI, Romania: IEEE, 2024: 1-6[2025-05-29]. https://ieeexplore.ieee.org/document/10805663/.
(ShaP C. The number of users of generative artificial intelligence products in China has reached 230 million.[EB/OL]. [2025-05-29].https://www.gov.cn/yaowen/liebiao/202411/content_6990362.htm)
[143]
LiJ J, GuanZ Y, WangJ, et al. Integrated image-based deep learning and language models for primary diabetes care[J]. Nature Medicine, 2024, 30(10): 2886-2896.
(Ranks among the international forefront and supports hierarchical medical diagnosis and treatment:the rare disease AI large model“Xiehe·Taichu” officially enters clinical application[EB/OL]. [2025-05-29]. https://www.pumch.cn/lys_details/40037.html.)