法發(fā)〔2022〕33號(hào)

最高人民法院

關(guān)于規(guī)范和加強(qiáng)人工智能司法應(yīng)用的意見(jiàn)

為深入學(xué)習(xí)貫徹黨的二十大精神，深入貫徹習(xí)近平法治思想，貫徹落實(shí)《中華人民共和國(guó)國(guó)民經(jīng)濟(jì)和社會(huì)發(fā)展第十四個(gè)五年規(guī)劃和2035年遠(yuǎn)景目標(biāo)綱要》和《新一代人工智能發(fā)展規(guī)劃》，推動(dòng)人工智能同司法工作深度融合，全面深化智慧法院建設(shè)，努力創(chuàng)造更高水平的數(shù)字正義，結(jié)合人民法院工作實(shí)際，制定本意見(jiàn)。

　　 一、指導(dǎo)思想

1.堅(jiān)持以習(xí)近平新時(shí)代中國(guó)特色社會(huì)主義思想為指導(dǎo)，深入貫徹習(xí)近平法治思想，堅(jiān)持司法為民、公正司法工作主線(xiàn)，加快推進(jìn)人工智能技術(shù)與審判執(zhí)行、訴訟服務(wù)、司法管理和服務(wù)社會(huì)治理等工作的深度融合，規(guī)范司法人工智能技術(shù)應(yīng)用，提升人工智能司法應(yīng)用實(shí)效，促進(jìn)審判體系和審判能力現(xiàn)代化，為全面建設(shè)社會(huì)主義現(xiàn)代化國(guó)家、全面推進(jìn)中華民族偉大復(fù)興提供有力司法服務(wù)。

　　 二、總體目標(biāo)

2.到2025年，基本建成較為完備的司法人工智能技術(shù)應(yīng)用體系，為司法為民、公正司法提供全方位智能輔助支持，顯著減輕法官事務(wù)性工作負(fù)擔(dān)，有效保障廉潔司法，提高司法管理水平，創(chuàng)新服務(wù)社會(huì)治理。到2030年，建成具有規(guī)則引領(lǐng)和應(yīng)用示范效應(yīng)的司法人工智能技術(shù)應(yīng)用和理論體系，為司法為民、公正司法提供全流程高水平智能輔助支持，應(yīng)用規(guī)范原則得到社會(huì)普遍認(rèn)可，大幅減輕法官事務(wù)性工作負(fù)擔(dān)，高效保障廉潔司法，精準(zhǔn)服務(wù)社會(huì)治理，應(yīng)用效能充分彰顯。

　　 三、基本原則

3.安全合法原則。堅(jiān)持總體國(guó)家安全觀(guān)，禁止使用不符合法律法規(guī)的人工智能技術(shù)和產(chǎn)品，司法人工智能產(chǎn)品和服務(wù)必須依法研發(fā)、部署和運(yùn)行，不得損害國(guó)家安全，不得侵犯合法權(quán)益，確保國(guó)家秘密、網(wǎng)絡(luò)安全、數(shù)據(jù)安全和個(gè)人信息不受侵害，保護(hù)個(gè)人隱私，促進(jìn)人機(jī)和諧友好，努力提供安全、合法、高效的智能化司法服務(wù)。

4.公平公正原則。堅(jiān)持遵循司法規(guī)律、服務(wù)公正司法，保證人工智能產(chǎn)品和服務(wù)無(wú)歧視、無(wú)偏見(jiàn)，不因技術(shù)介入、數(shù)據(jù)或模型偏差影響審判過(guò)程和結(jié)果的公正，同時(shí)尊重不同利益訴求，能夠根據(jù)司法需求公平提供合理可行方案，充分照顧困難群體、特殊群體，使其在司法活動(dòng)中獲得必要幫助，實(shí)現(xiàn)智能化司法服務(wù)對(duì)各類(lèi)用戶(hù)的普適包容和機(jī)會(huì)均等。

5.輔助審判原則。堅(jiān)持對(duì)審判工作的輔助性定位和用戶(hù)自主決策權(quán)，無(wú)論技術(shù)發(fā)展到何種水平，人工智能都不得代替法官裁判，人工智能輔助結(jié)果僅可作為審判工作或?qū)徟斜O(jiān)督管理的參考，確保司法裁判始終由審判人員作出，裁判職權(quán)始終由審判組織行使，司法責(zé)任最終由裁判者承擔(dān)。各類(lèi)用戶(hù)有權(quán)選擇是否利用司法人工智能提供的輔助，有權(quán)隨時(shí)退出與人工智能產(chǎn)品和服務(wù)的交互。

6.透明可信原則。堅(jiān)持技術(shù)研發(fā)、產(chǎn)品應(yīng)用、服務(wù)運(yùn)行的透明性，保障人工智能系統(tǒng)中的司法數(shù)據(jù)采集管理模式、法律語(yǔ)義認(rèn)知過(guò)程、輔助裁判推定邏輯、司法服務(wù)互動(dòng)機(jī)制等各個(gè)環(huán)節(jié)能夠以可解釋、可測(cè)試、可驗(yàn)證的方式接受相關(guān)責(zé)任主體的審查、評(píng)估和備案。司法人工智能產(chǎn)品和服務(wù)投入應(yīng)用時(shí)，應(yīng)當(dāng)以便于理解的方式說(shuō)明和標(biāo)識(shí)相應(yīng)的功能、性能與局限，確保應(yīng)用過(guò)程和結(jié)果可預(yù)期、可追溯、可信賴(lài)。

7.公序良俗原則。堅(jiān)持將社會(huì)主義核心價(jià)值觀(guān)融入司法人工智能技術(shù)研發(fā)、產(chǎn)品應(yīng)用和服務(wù)運(yùn)行全過(guò)程，保證人工智能司法應(yīng)用不得違背公序良俗，不能損害社會(huì)公共利益和秩序，不能違背社會(huì)公共道德和倫理，健全風(fēng)險(xiǎn)管控、應(yīng)急處置和責(zé)任查究機(jī)制，防范化解人工智能司法應(yīng)用中可能產(chǎn)生的倫理道德風(fēng)險(xiǎn)。

　　 四、應(yīng)用范圍

8.加強(qiáng)人工智能全流程輔助辦案。支持證據(jù)指引與審查、法律法規(guī)推送、類(lèi)案推送、全案由裁判輔助、法律文書(shū)輔助生成、法律文書(shū)輔助審查等智能化應(yīng)用，促進(jìn)裁判尺度統(tǒng)一，保障司法公正，維護(hù)司法權(quán)威。

9.加強(qiáng)人工智能輔助事務(wù)性工作。支持電子卷宗自動(dòng)分類(lèi)歸目、案件信息自動(dòng)回填、案件繁簡(jiǎn)分流、送達(dá)地址及方式自動(dòng)推薦、司法活動(dòng)筆錄自動(dòng)生成、執(zhí)行財(cái)產(chǎn)查控輔助、電子卷宗自動(dòng)歸檔等智能化應(yīng)用，降低各類(lèi)人員工作負(fù)擔(dān)，提高司法效率。

10.加強(qiáng)人工智能輔助司法管理。支持案件裁判偏離度預(yù)警、終本案件核查、不規(guī)范司法行為自動(dòng)巡查、廉潔司法風(fēng)險(xiǎn)防控等智能化應(yīng)用，提升司法管理質(zhì)效，保障廉潔司法。

11.加強(qiáng)人工智能服務(wù)多元解紛和社會(huì)治理。支持司法資源推薦、訴訟和調(diào)解咨詢(xún)問(wèn)答、訴訟預(yù)期輔助評(píng)估、社會(huì)治理風(fēng)險(xiǎn)預(yù)警與輔助決策等智能化應(yīng)用，為化解社會(huì)矛盾、服務(wù)社會(huì)治理提供新的途徑和方式。

12.不斷拓寬人工智能司法應(yīng)用場(chǎng)景和范圍。結(jié)合人工智能技術(shù)創(chuàng)新進(jìn)程和人民法院改革發(fā)展實(shí)踐，積極探索訴訟服務(wù)、審判執(zhí)行、司法管理和服務(wù)社會(huì)治理等領(lǐng)域的重大應(yīng)用場(chǎng)景，不斷拓展新的應(yīng)用范圍。

　　 五、系統(tǒng)建設(shè)

13.加強(qiáng)人工智能應(yīng)用頂層設(shè)計(jì)。按照人民法院信息化建設(shè)發(fā)展規(guī)劃部署，設(shè)計(jì)完善智慧法院人工智能相關(guān)信息系統(tǒng)體系架構(gòu)和技術(shù)標(biāo)準(zhǔn)體系，豐富拓展人工智能司法應(yīng)用場(chǎng)景，建立健全人工智能系統(tǒng)信息安全和運(yùn)維保障制度，指導(dǎo)和規(guī)范各級(jí)人民法院人工智能系統(tǒng)建設(shè)。

14.加強(qiáng)司法數(shù)據(jù)中臺(tái)和智慧法院大腦建設(shè)。加快推進(jìn)司法數(shù)據(jù)庫(kù)、數(shù)據(jù)服務(wù)平臺(tái)、司法知識(shí)庫(kù)、人工智能引擎、知識(shí)服務(wù)平臺(tái)和司法區(qū)塊鏈平臺(tái)等系統(tǒng)的建設(shè)和集成，打造實(shí)體化司法數(shù)據(jù)中臺(tái)和智慧法院大腦，為面向各類(lèi)業(yè)務(wù)的人工智能司法應(yīng)用提供核心驅(qū)動(dòng)。

15.加強(qiáng)司法人工智能應(yīng)用系統(tǒng)建設(shè)。圍繞人民法院司法活動(dòng)典型業(yè)務(wù)場(chǎng)景，以提升智能化水平為主線(xiàn)，促進(jìn)司法數(shù)據(jù)中臺(tái)和智慧法院大腦與智慧服務(wù)、智慧審判、智慧執(zhí)行和智慧管理等業(yè)務(wù)應(yīng)用系統(tǒng)融合集成，不斷提供滿(mǎn)足司法業(yè)務(wù)需求、符合先進(jìn)技術(shù)發(fā)展方向的司法人工智能產(chǎn)品和服務(wù)。

16.加強(qiáng)司法人工智能關(guān)鍵核心技術(shù)攻關(guān)。依托國(guó)家重點(diǎn)工程、科研項(xiàng)目和科技創(chuàng)新平臺(tái)，組織產(chǎn)學(xué)研優(yōu)勢(shì)力量，發(fā)揮學(xué)科交叉催化劑作用，針對(duì)面向司法語(yǔ)境的大規(guī)模預(yù)訓(xùn)練語(yǔ)言模型及其應(yīng)用、多模態(tài)司法大數(shù)據(jù)高效處理方法、司法數(shù)據(jù)驅(qū)動(dòng)與知識(shí)引導(dǎo)相結(jié)合的深度神經(jīng)網(wǎng)絡(luò)模型構(gòu)建與樣本學(xué)習(xí)方法、基于法律知識(shí)增強(qiáng)的可解釋檢索和推理模型、面向司法效能提升的人機(jī)交互范式、基于新一代人工智能的審判輔助系統(tǒng)等關(guān)鍵核心技術(shù)集智攻關(guān)，為司法人工智能系統(tǒng)建設(shè)提供牽引和支撐。

17.加強(qiáng)基礎(chǔ)設(shè)施建設(shè)和安全運(yùn)維保障。根據(jù)司法人工智能對(duì)算力、通信和服務(wù)能力的需求，科學(xué)合理地規(guī)劃和建設(shè)通信網(wǎng)絡(luò)、計(jì)算存儲(chǔ)、通用終端設(shè)備和專(zhuān)用信息化設(shè)施等信息基礎(chǔ)設(shè)施，強(qiáng)化網(wǎng)絡(luò)安全、數(shù)據(jù)安全和個(gè)人信息保護(hù)能力，完善人工智能運(yùn)行維護(hù)機(jī)制，為人工智能司法應(yīng)用提供必要的保障條件。

　　 六、綜合保障

18.提高思想認(rèn)識(shí)，加強(qiáng)組織領(lǐng)導(dǎo)。高度重視人工智能應(yīng)用對(duì)司法為民、公正司法的重要意義，以智慧法院新一代人工智能示范應(yīng)用為契機(jī)，找準(zhǔn)工作結(jié)合點(diǎn)、切入點(diǎn)，把握發(fā)展規(guī)律，爭(zhēng)取資金支持，注重宣傳培訓(xùn)，引導(dǎo)干警充分參與，努力推動(dòng)司法人工智能應(yīng)用取得突破。

19.促進(jìn)協(xié)同創(chuàng)新，保護(hù)知識(shí)產(chǎn)權(quán)。加強(qiáng)司法大數(shù)據(jù)質(zhì)量管控，完善跨部門(mén)、跨層級(jí)、跨業(yè)務(wù)的司法數(shù)據(jù)協(xié)同共享和智能化服務(wù)共建共享機(jī)制，支持司法人工智能科技創(chuàng)新和專(zhuān)利、軟件著作權(quán)申報(bào)，切實(shí)保護(hù)相關(guān)知識(shí)產(chǎn)權(quán)。

20.加強(qiáng)安全保障，防范安全風(fēng)險(xiǎn)。加強(qiáng)司法數(shù)據(jù)分類(lèi)分級(jí)管理，強(qiáng)化重要數(shù)據(jù)和敏感信息保護(hù)，完善司法數(shù)據(jù)安全共享和應(yīng)用模式，通過(guò)司法人工智能倫理委員會(huì)等機(jī)制，綜合采用倫理審核、合規(guī)審查、安全評(píng)估等方式，防范化解人工智能應(yīng)用過(guò)程中的安全風(fēng)險(xiǎn)。

最高人民法院

2022年12月8日

The Supreme People’s Court

The Opinions on Regulating and Strengthening the Applications of Artificial Intelligence in the Judicial Fields

In order to thoroughly study and implement Xi Jinping Thought on the Rule of Law and the spirit of the 20th Communist Party of China (CPC) National Congress, and to enforce the Outline of the 14th Five-Year Plan (2021-2025) for National Economic and Social Development and the Long-Range Objectives Through the Year 2035 and the New-generation Artificial Intelligence Development Plan, the Supreme People’s Court (SPC), aligning with the practice of the people’s courts, is now publishing the Opinions. The Opinions aim to promote the in-depth integration of artificial intelligence with judicial work, deepen the construction of smart courts, and strive to achieve a higher level of digital justice.

　　 I. Guiding Principles

1. People’s courts must follow the guidance of Xi Jinping Thought on Socialism with Chinese Characteristics for a New Era, thoroughly implement Xi Jinping Thought on the Rule of Law, and uphold justice for the people and judicial impartiality as our cardinal work. People’s courts shall advance in-depth integration of artificial intelligence with adjudication and enforcement, litigation service, court management, as well as social governance facilitation, regulate yet pursue effective application of artificial intelligence in the judicial fields, to accelerate the modernization of the judicial system and judicial capability, and to serve the building China into a modern socialist country in all respects and promoting the great rejuvenation of the Chinese nation at all fronts with judicial safeguards.

　　 II. Overall Objectives

2. People’s courts shall, by the year 2025, construct an improved functional system for the application of artificial intelligence in the judicial field, the function of which is to provide all-round intelligent support for serving the people and justice, and to effectively alleviate the high administrative workload of judges, thus achieving improvement in anti-corruption and court management, and innovation in facilitating social governance. People’s courts shall, by the year 2030, build an application and theoretical system for the utilization of artificial intelligence in the judicial field with model rules and demonstration effects, to provide whole-process high-level intelligent support for serving the people and justice, with the normsand principles being widely acknowledged, the administrative workload of judges being significantly reduced, providing effective and adequate safeguards for anti-corruption, precise facilitation for social governance, and achieving full application effectiveness.

　　 III.General Principles

3. Principle of Security and Legality. Pursuing a holistic approach to national security, forbidding the usage of illegal AI technologies and products. Judicial AI products and services shall be legally developed, deployed and operated, and shall not be detrimental to national security and legal interests. It shallsecure state secrets, network security, data security and personal information free from infringement, protect personal privacy, promote a harmonious and friendly interface between the user and AI, and provide safe, legitimate, and efficient intelligent judicial services.

4. Principle of Fairness and Justice. Following fundamental judicial rules, serving judicial fairness, ensuring AI products and services are free from discrimination and prejudice. The fairness of the processes and outcomes of trials shall not be impaired by the introduction of technology, either model nor data deviations. Meanwhile, respecting diverse interests, offering fair, reasonable and feasible solutions based on judicial demands, adequately providing necessary assistance to communitiesin difficulties and people with special needs to participate in judicial activities. Intelligent judicial services shall achieve universal inclusion of all groups of users with equal opportunities.

5. Principle of Supporting Adjudication. Affirming the supportive role of AI in adjudication, and the user’s rights to decision-making. AI shall not make judicial decision substituting for the judge in any case, disregarding technology advancement. The results from AI shall be for supplemental references only, for adjudication or judicial supervision and management. Ensuring all judicial decision are made by the judicature, all judicial powers are administered by adjudicative authorities, and all judicial accountability ultimately falls on the decision-maker. All users have the rights to decide whether or not to use judicial AI assistance and the rights to terminate their interface with AI products and services.

6. The Principle of Transparency and Credibility. Ensuring the transparency of technology development, product application and service operation. Ensuring all links of AI systems, including the collection and management patterns of judicial data, the process of legal cognitive semantics, and the logic of assisting judicial presumptions would accept examination, evaluation and registration with relevant authoritative entities with interpretability, testability and verifiability. Corresponding features, capabilities and limitations shall be instructed and identified in a manner that can be easily understood when any judicial AI products and services are used, to ensurethat the procedure and outcome of applications are predictable, traceable and credible.

7. The Principle of Abiding By PublicOrder and Good Customs. Infusing the Core Socialist Values into the whole process of technology development, product application and service operation of judicial AI. The application of judicial AI shall not endanger public order and good customs, shall not damage public interests and social order, and shall not violate public morals and ethics. A mechanism of risk management, emergency response and responsibility investigation shall be established to avoid and resolve possible moral and ethical risks from the application of judicial AI.

　　 IV. Application Scope

8. Enhancing AI-assisted case-handling in whole process. People’s courts shall support the development of AI applications on evidence guidance and review, smart push of laws, regulations and similar cases, adjudication assistance for all causes of actions, AI-assistedlegal documentsgeneration and review, to further advance the uniformity of adjudication, secure judicial justice and safeguard judicial authority.

9. Enhancing AI-assisted complementation of administrative work. People’s courts shall support the development of AI applications for e-filesclassification and categorization, case informationcrawling, automatic diversion of complex and simple cases, automatic recommendation of service addresses and methods, automatic generation of judicial records, AI-assisted property investigation and seizure, automatic electronic casefiling and other related aspects, to reduce the administrative workload of all types of judicial personnel and improve judicial efficiency.

10. Enhancing AI-assisted judicial management. People’s courts shall support the development of AI applications forwarning deviationof adjudicative criteria, review of the procedural-terminated enforcement cases, inspection of judicial irregularities, prevention and control of judicial corruption risk and other related aspects, to optimize the quality and efficiency ofjudicial management and maintain judicial integrity.

11. Enhancing AI services for diversified dispute resolution and social governance. People’s courts shall support the development of AI applications for judicial resolution recommendation, litigation and mediation consultation or Q&A, AI-assisted litigation prediction, early warning of social governance risk, AI-assisted decision-making and other related aspects, to provide new methods for resolving social conflicts and serving social governance.

12. Expanding the AI application scope in judicial fields. People’s courts shall combine the innovation of AI technology with the practice of judicial reform, exploring AI application scenarios in the fields of litigation service, adjudication, enforcement, judicial management and social governance, and expanding new application scope within different fields.

　　 V. System Construction

13. Strengthening the top-level design of AI application. According to the Plan for Information Construction at People’s Courts, the Supreme People’s Court shall build and improve the structure of the Smart AI Court Information System and its technical standard. TheSupreme People’s Court shall broaden the application scenarios of AI Judiciary while establishing and improving the information security, operation and maintenance mechanism of the AI System, and shall guide and regulate the AI system construction of people’s courts at each level.

14. Strengthening the construction ofjudicial data centers and smart court brains.People's courts shall expedite the construction and integration of systems, including judicial databases, data service platforms, judicial knowledge bases, artificial intelligence engines, knowledge service platforms and judicial blockchain platforms. These measures aim to virtualize and enhance the judicial data centers and smart court brains, providing the core driving force for AI applications on judicial services.

15. Strengthening the construction of judicial AI application system. With the main focus being improving the AI performance, People’s courts shall promote the integration of AI application systems including judicial data centres, smart court brains, smart services, smart adjudication, smart enforcement, smart management and other typical judicial scenarios. People’s courts shall continuously provide advanced judicial AI products and services to meet the growingjudicial needs.

16. Reinforcing research on key and core technologies of judicial AI. Based on the strength of national significant projects, scientific research programs and technology innovation platforms, the SPC shall organize leading research forces from industries, academies and laboratories, enhancing the catalyst effect of interdisciplinary study. The research shall focus on pre-trained language models and application regarding judicial context, high efficiency multimodal judicial data processing, judicial data-driven and knowledge-oriented deep neural networks model and transfer learning, legal knowledge-enhanced interpretable retrieval and reasoning model, human-computer interaction aiming at improving judicial efficiency, adjudication support system based on new generation AI and other core technologies, thus promoting and upholding the construction of judicial AI system.

17. Strengthening the construction of infrastructure, security operation and maintenance. Accommodating the needs of judicial AI on computing power, communication and service capacity, People’s courts shall plan and construct information infrastructure in a reasonable scientific manner, including communication network, computing storage, terminal equipment，and specialized information facility. These plans and construction shall strengthen internet security and data security, further protect personal information and improve AI’s operation and maintenance mechanism, providing vital protection and condition for judicial AI applications.

　　 VI. Comprehensive Supports

18. Guiding the Development with Higher Awarenessand Deeper Understanding. By attaching great importance to judicial AI applications in realizing justice for the people and judicial impartiality,People’s courts shall take the opportunity of Smart Courts construction, one of the selected AI model application scenarios, to identify valuable conjunction and connection points, and grasp the laws governing development. More efforts shall be devoted to attracting financial support, enhancing publicity and training, encouraging more judicial personnel to fully participate, andfurther promoting breakthroughs in judicial AI applications.

19. Promoting Collaborative Innovation to Protect Intellectual Property Rights. People’s courts shall increase the quality management of judicial Big-Data supplies to fully exploit and develop systems with the features of inter-departmental, cross-level and cross-service that share collaborative judicial data and intelligent judicial service. People’s courts shall support judicial AI innovations, encouragerelated patent applications or software copyright registrations, providing sufficient protection of intellectual property rights.

20. Enhancing Security Management to MitigateCybersecurity Risks. People’s courts shall promote categorized and hierarchical management of judicial data, enhance the protection of crucial data and sensitive information, and secure the safety of judicial data developing and sharing. Through mechanisms such as Judicial AI Ethics Council, People’s courts shall comprehensively adopt methods,including ethical reviews, compliance reviews, and security assessments, to prevent and mitigate cybersecurity risks in judicial AI applications.

責(zé)任編輯：藺弦弦