Supreme People’s Court Issues Provisions on Several Issues Concerning the Trial of Civil Disputes Involving Criminal Offences

Date of issuance: 21 September 2026
Effective date: 22 September 2026

On 20 September 2026, the Supreme People’s Court issued these Provisions, a judicial interpretation comprising 24 articles that came into force on 22 September 2026. The Provisions coordinate the substantive rules and procedures applicable to civil and criminal matters. Key provisions include:

1. Improving coordination between civil and criminal proceedings: The Provisions determine the applicable procedure based on whether the civil and criminal cases arise from the same basic facts. Where they do, criminal proceedings generally take precedence. For related proceedings concerning different facts, the Provisions specify six circumstances in which the civil and criminal cases may proceed separately and in parallel. In such cases, the Court may stay the civil proceedings only if the determination of its basic facts necessarily depends on the outcome of the criminal proceedings.
2. Harmonising fact-finding and evidential rules in overlapping civil and criminal cases: Facts established by a final criminal judgment will be binding in related civil proceedings unless rebutted by sufficient evidence. Confessions, witness testimony, and other evidence that is neither relied upon nor excluded in the criminal proceedings may be examined under civil procedure rules. Facts failing to meet the criminal standard of proof may still be established in accordance with the civil standard.
3. Clarifying contract validity and liability where the act of entering into the contract constitutes a criminal offence: The fact that entering into a contract constitutes a criminal offence does not necessarily render the contract invalid. Its validity must be determined under the relevant provisions of the Civil Code. The Provisions set out the applicable rules for different circumstances.
4. Improving the property handling and right protection mechanisms in overlapping civil and criminal proceedings: The Provisions set out how civil compensation is to be coordinated with the recovery and restitution of property in criminal proceedings. Any property returned or compensation paid through criminal recovery, restitution, or other measures, must be deducted when the subsequent civil judgment is rendered or enforced. The Provisions also allow eligible enterprises involved in criminal offences to enter bankruptcy proceedings and provide for coordination between the recovery of unlawful proceeds and the distribution of the bankruptcy estate.
5. Strengthening remedies available to the parties: The parties may still initiate civil proceedings where a criminal case is not accepted, withdrawn or not prosecuted, or where the criminal proceedings result in an acquittal or a finding that no criminal liability should be imposed.

Source: Supreme People’s Court
https://mp.weixin.qq.com/s/hhive_k7NNSPL3JswwGdUw
https://mp.weixin.qq.com/s/JyR2Mswe1Aj3ta_-yqlpJg

Supreme People’s Court Issues Opinions on the Lawful Adjudication of AI-Related Disputes

Document No.: Fa Fa [2026] No. 10
Date of issuance: 7 September 2026
Effective date: 7 September 2026

The Supreme People’s Court’s first guidance on adjudicating AI-related cases, the Opinions on the Lawful Adjudication of AI-Related Disputes (the Opinions), comprises five parts and 24 articles. Based on existing laws, including the Civil Code, the Copyright Law, the Personal Information Protection Law, the Cybersecurity Law and the Civil Procedure Law, it addresses AI-related infringement, intellectual property, personality rights, consumer rights, product liability, regulation of self-driving automobiles, litigation procedures and criminal liability. Key provisions on intellectual property protection, personality rights and “AI hallucinations” include:

1. Clarifying liability for copyright infringement by AI-generated content: The fact that content is AI-generated does not, in itself, preclude copyright infringement. When allocating liability, the Court must assess each party’s degree of control, duty of care and fault during AI generation, taking into account the characteristics of the AI technology and business model, the respective roles of the developer, service provider and user, the party that supplied the training data, the precautions taken, and the benefits obtained.
2. Clarifying the burden of proof in copyright disputes involving AI-generated content: A rights holder claiming copyright infringement by an AI developer or service provider must present prima facie evidence that the disputed content was generated by the AI system in question and that it is substantially similar to the copyright work. To contest infringement, the developer or service provider must provide evidence of matters within its control, including the source of the training data, the training process, the model’s operation, and the relevant scientific basis where necessary. An AI user who knows or ought to know of an earlier work, but uses AI to generate substantially similar content is liable for infringement unless a defence, such as fair use, applies.
3. Regulating unfair competition involving AI-generated content: Where a business operator uses AI-generated text, images, videos or virtual characters for the purpose of passing off, distributing false advertising, or fabricating traffic figures or positive user reviews through AI face-swapping or similar means, and thereby harms the lawful rights and interests of others, misleads consumers or disrupts market competition, the Court should rule, in accordance with law, that the conduct amounts to unfair competition, including passing off or false advertising.

5. Copyrightability of AI-generated content and the use of works for model training remain unaddressed: The Supreme People’s Court noted that there remains considerable disagreement over whether AI-generated content qualifies for copyright protection and how to treat the unauthorised use of others’ works to train large AI models. The Opinions therefore do not address these two issues, leaving the relevant adjudication rules to be developed in light of technological advances and judicial practice.
6. Regulating the use of AI to impersonate celebrities in product promotions: Where AI is used to impersonate a celebrity in promoting goods, thereby infringing the rights of others and engaging in fraud, the Court shall hold relevant parties liable for the infringement and uphold consumers’ claims for punitive damages.
7. Strengthening the protection of personality rights, including rights relating to likeness, name, voice, reputation and privacy: Using AI, without an individual’s consent, to process personal information such as name or likeness in order to create and use or disclose an identifiable virtual representation, infringes the individual’s rights relating to name, likeness and other personality rights. Using an individual’s voice as training data to generate an identifiable synthetic voice that imitates timbre, intonation and the person’s speaking style will infringe voice rights unless consent has been obtained. The Opinions also regulate privacy rights infringements involving the use of AI to track and analyse publicly available personal information for the purpose of obtaining, disclosing or publicising private information, as well as practices such as ‘doxxing’ and ‘human flesh searches’. Eligible victims may apply for an injunction against infringement of personality rights.
8. Drawing the line between ‘AI hallucinations’ and deliberate prompting of infringing content: Where generative AI independently produces content that infringes personality rights, the service provider should be held liable if it fails to act promptly upon notice from the rights holder. Where a user deliberately induces the generation of infringing content and causes harm, the user shall be held liable. The service provider shall bear a corresponding liability if, upon notice, it fails to stop the generation, block the relevant prompts or take other necessary measures.
9. Addressing the misuse of AI in litigation: Parties to litigation should verify the evidence, pleadings and case search reports they submit, disclose any use of AI assistance, and ensure that the content is authentic and accurate. Where AI is used to fabricate case facts, bring sham proceedings or falsify evidence, the Court may dismiss the claims and impose a fine or detention. Criminal liability should apply where the conduct constitutes a criminal offence.

The Opinions also address the use of lawfully disclosed personal information for model training within reasonable limits, algorithmic discrimination, AI-generated celebrity endorsements, defects in AI products, liability for self-driving automobile accidents, open-source software liability, the grant and validity of patents for AI-related inventions, the performance of technology contracts and rules governing data use. They further call for improved mechanisms for determining technical facts, reviewing evidence, providing judicial guidance and facilitating interdepartmental coordination.

Source: China Court Network

https://www.court.gov.cn/zixun/xiangqing/511101.html

最高法发布《关于审理涉及刑事犯罪的民事纠纷案件若干问题的规定》

发文日期：2026-09-21

生效日期：2026-09-22

最高人民法院于2026年9月20日发布《关于审理涉及刑事犯罪的民事纠纷案件若干问题的规定》（下称《规定》），该司法解释自2026年9月22日起施行。《规定》共24条，重点针对如何衔接协调民刑两大领域的实体规则和诉讼程序作出解释，具体包括以下内容：

• 完善“民刑并行”程序衔接机制：《规定》以刑事与民事案件所基于的基本事实判断审理规则。基于同一事实的，原则上适用先刑后民规则，但如果两个诉讼虽然存在关联，但并非涉及同一事实的情况下，《规定》列举了六种允许“民刑并行”、分别审理的情形。此时，仅在民事案件基本事实的认定必须以刑事案件审理结果为依据时，法院方可裁定中止民事诉讼。
• 统一民刑交叉案件中的事实认定与证据规则：生效刑事裁判已确认的事实在相关民事诉讼中原则上无需重复举证，但有相反证据足以推翻的除外；刑事程序中未被作为定案依据且未被排除的供述、证言等证据材料，可以依照民事诉讼规则进行质证和审查；因未达到刑事诉讼证明标准而未被刑事裁判认定的事实，只要达到民事诉讼证明标准，仍可在民事案件中依法认定。
• 完善涉刑合同效力及责任承担规则：涉刑合同并不当然无效，其效力应依据《民法典》相关规定判断。《规定》区分不同情形明确裁判规则。
• 健全民刑程序衔接下的财产处置和权益保护机制：明确民事赔偿与刑事追缴、退赔之间的衔接规则，对已通过刑事追缴、退赔等方式返还或赔偿的部分自民事裁判和执行中予以扣减；同时设立“刑破并行”规则，允许符合条件的涉罪企业进入破产程序，并统筹协调违法所得追缴与破产财产分配权益。
• 强化当事人权利救济：对于不立案、撤案、不起诉、宣告无罪或者不负刑事责任等情形，当事人仍可依法主张民事责任。

资料来源：最高人民法院  2026-09-21

新闻链接：https://mp.weixin.qq.com/s/hhive_k7NNSPL3JswwGdUw; https://mp.weixin.qq.com/s/JyR2Mswe1Aj3ta_-yqlpJg

最高法发布《关于依法审理涉人工智能纠纷案件的意见》

发文字号：法发〔2026〕10号

发文日期：2026-09-07

生效日期：2026-09-07

《意见》是最高人民法院发布的首部涉人工智能司法裁判规则文件，共分为5个部分、24条，依据《民法典》《著作权法》《个人信息保护法》《网络安全法》《民事诉讼法》等现行法律，对涉人工智能侵权、知识产权、人格权益、消费者权益、产品责任、自动驾驶、诉讼程序及刑事责任等问题提出裁判指引。《意见》中与知识产权保护、人格权益及“AI幻觉”相关的核心要点如下：

1. 明确人工智能生成内容侵害著作权的责任认定规则：人工智能生成内容被诉侵害他人著作权的，相关被诉主体不得仅以被诉内容由人工智能生成为由主张免责。法院应当根据人工智能技术及商业模式的特点、开发者、服务提供者和使用者在内容生成过程中发挥的作用、训练数据由何方提供、各方采取的必要措施及获利情况等因素，判断相关主体的控制能力、注意义务、主观过错及过错程度，并据此合理划分侵权责任。
2. 细化人工智能生成内容著作权纠纷中的举证责任：权利人向人工智能服务开发者或者提供者主张著作权侵权的，应当初步证明被诉侵权内容由涉案人工智能生成，并与权利作品构成实质性相似。开发者或者提供者提出不侵权抗辩的，应当就模型训练数据来源、训练过程及运行模式等由其掌握的事实举证，必要时还应提供相关科学理论依据。人工智能使用者知道或者应当知道在先作品，仍利用人工智能生成与该作品实质性相似的内容，且不存在合理使用等免责事由的，应当认定构成侵权。
3. 规制利用人工智能生成内容实施的不正当竞争行为：经营者利用人工智能生成图文、视频或虚拟人物形象，实施仿冒、虚构流量和用户好评，或者通过AI换脸等方式进行虚假宣传，损害他人合法权益、误导消费者或者扰乱市场竞争秩序的，人民法院应当依法认定相关行为构成仿冒侵权、虚假宣传等不正当竞争行为。
4. 暂未规定人工智能生成内容的可版权性及模型训练使用作品问题：最高人民法院表示，关于人工智能生成内容能否获得著作权保护，以及未经许可使用他人作品训练人工智能大模型应如何定性，目前仍存在较大分歧。因此，《意见》暂未对上述两个问题作出规定，相关裁判规则有待结合技术发展及司法实践进一步探索。
5. 依法规制利用人工智能“仿冒名人带货”：利用人工智能仿冒名人进行商品推广，侵害他人合法权益并构成欺诈的，人民法院应当依法认定相关主体承担侵权责任；消费者主张惩罚性赔偿的，人民法院依法予以支持。
6. 加强对肖像、姓名、声音、名誉和隐私等人格权益的保护：未经自然人同意，利用人工智能处理其姓名、肖像等信息，生成能够识别该自然人的虚拟数字形象并予以使用或者公开的，构成对姓名权、肖像权等人格权益的侵害；未经同意将自然人的声音作为训练语料，模仿其音色、语调及发音风格生成可识别的合成人声的，构成对声音权益的侵害。《意见》亦对利用人工智能追踪、分析个人公开信息以获取、泄露或者公开私密信息，以及“网络开盒”“人肉搜索”等侵犯隐私权的行为作出规范，并明确符合条件的受害人可以申请人格权侵害禁令。
7. 明确“AI幻觉”及恶意诱导生成侵权内容的责任边界：生成式人工智能自动生成侵害人格权益的内容，经权利人通知后，服务提供者未及时采取停止生成等必要措施的，应当依法承担侵权责任。用户通过输入侵权提示词等方式恶意诱导生成侵权内容并造成损害的，由用户承担侵权责任；服务提供者接到通知后未及时停止生成相关内容、屏蔽生成指令或者采取其他必要措施的，也应依法承担相应责任。
8. 不得以“AI幻觉”或者技术中立为由免除诉讼责任：诉讼参与人应当核实其提交的证据、诉讼文书和案例检索报告，说明使用人工智能辅助的情况，并对相关内容的真实性、准确性负责。利用人工智能捏造案件事实、进行虚假诉讼或者伪造证据的，法院可以依法驳回诉讼请求并予以罚款、拘留；构成犯罪的，依法追究刑事责任。

除上述内容外，《意见》还涉及在合理范围内使用已合法公开的个人信息进行模型训练、算法差别待遇、利用人工智能仿冒名人进行商业推广、人工智能产品缺陷、自动驾驶事故责任、开源软件责任、人工智能相关发明创造的专利授权确权、技术合同履行及数据使用规范等问题，并要求完善技术事实查明、证据审查、审判指导及部门协作机制。

资料来源：中国法院网  2026-09-07

新闻链接：https://www.court.gov.cn/zixun/xiangqing/511101.html

Kaifu Primary People’s Court, Changsha: AI-Generated Image Not Copyrightable Where Prompts and Edits Lack Originality

Date: 24 September 2026

In May 2025, a third party used an AI image-generation model to create an image of a Dragon Boat Festival sachet. The user selected only a 1:1 aspect ratio and 1K resolution and entered prompts describing the sachet, its colours and a Chinese-style aesthetic. The user then removed repeated text from the bottom of the image using Photoshop.  It assigned all rights in the image to the Plaintiff, who proceeded to register it as a copyright work. When a kindergarten in Changsha subsequently used the image on its official WeChat account, the Plaintiff brought an action for infringement of the right of communication through information networks, seeking RMB 10,000 (approx. US$ 1,400) compensation for economic loss as well as reasonable enforcement expenses.

The Changsha Kaifu Primary People’s Court held that the use of an AI tool to generate material does not, in itself, determine whether the resulting content qualifies as a copyright work. The decisive question is whether the generation process reflects original intellectual input by a natural person, assessed at three levels: front-end conception, control over generation and back-end processing. In assessing control over generation, a Court should consider the user’s control over the specific expression, the extent to which the prompts correspond to the final image and whether the user’s choices played a predominant role.

In this case, the prompts contained only general descriptions of the subject matter, theme, colours and style. They reflected no personal choices regarding the sachet’s shape, the image’s composition, the arrangement of its elements, or the use of light and shadow. The aspect ratio and resolution were merely technical settings, while the subsequent removal of repeated text was a technical correction that added no original expression. The Court therefore held that the image did not qualify as a work under copyright law and dismissed all the Plaintiff’s claims. Neither party appealed, and the judgment has taken effect.

Source: China News Service, Hunan

https://www.hn.chinanews.com.cn/news/sxdt/2026/0924/534232.html

长沙开福法院：提示词及后续修改缺乏独创性时，涉案AI生成图片不构成作品

日期：2026-09-24

2025年5月，案外人使用AI图像生成模型制作一张端午节香囊图片，仅设置1:1比例和1K清晰度，并输入关于端午节香囊、色彩及中式风格的提示词，随后使用Photoshop删除图片下方重复的文字。案外人之后将图片著作财产权转让给原告并完成作品登记。长沙某幼儿园在微信公众号中使用该图片后，原告以侵害信息网络传播权为由提起诉讼，索赔经济损失及维权开支1万元。

长沙市开福区人民法院认为，判断AI生成物是否构成作品，关键不在于是否使用人工智能工具，而在于生成过程是否体现自然人的独创性智力投入。对此，可从前端构思、生成控制和后端加工三个层次分别审查；其中，生成控制环节应重点考察控制行为是否直接指向具体表达、提示词与最终画面是否相对应，以及使用者的选择是否占据主导地位。

本案提示词仅对生成对象、题材、色彩和风格提出概括性要求，未对香囊造型、画面构图、元素排列及光影层次等具体表达作出个性化选择，图片比例和清晰度也仅属一般技术参数。后续删除重复文字仅为技术性修正，未形成新的独创性表达。因此，法院认定涉案AI生成图片不构成著作权法意义上的作品，驳回原告全部诉讼请求。双方均未上诉，判决现已生效。

资料来源：中国新闻网湖南  2026-09-24

新闻链接：https://www.hn.chinanews.com.cn/news/sxdt/2026/0924/534232.html

Guangzhou Intellectual Property Court: ‘Jiaodian’ Ordered to Pay ‘Bananain’ RMB 5 Million for Trade Mark and Online Store Decoration Infringement

Date: 23 September 2026

Sanliren Company owns the registered trade marks in the ‘蕉内’ (Bananain) series. As a result of long-term operation, its official flagship online store has developed a distinctive overall image. You Company and its sole shareholder Wu applied to register ‘蕉点’ ‘JIAODIAN蕉点’ (pinyin version) and other relevant marks in relation to the same types of product that Sanliren Company sold under its ‘Bananain’ marks. You Company manufactured and sold products bearing the ‘JIAODIAN蕉点’ and related marks; Ling Company and Run Company promoted and sold the products through their online stores. Despite invalidation of the ‘Jiaodian’ marks, upheld by final administrative judgments, You Company continued to apply for the registration of similar marks and to sell products bearing those marks. Sanliren Company brought an action before the Guangzhou Yuexiu Primary People’s Court.

At first instance, the Court held that the exclusive rights in the invalidated ‘Jiaodian’ marks were deemed never to have existed. Their use therefore lacked a legitimate basis and infringed Sanliren Company’s trade marks. The Defendants’ online stores also closely resembled the official ‘Bananain’ flagship store in their background colours, promotional slogans, page layouts and product displays, causing actual consumer confusion. The Court found that the Defendants had imitated the overall decoration of an influential online store and thereby engaged in unfair competition.

In assessing damages, the Court considered You Company’s role as the source of the infringing products, the duration and scale of the infringement, the substantial sales generated, and its imitation of both the brand identifiers and store decoration, which demonstrated a clear intention to exploit the reputation of ‘Bananain’. The Court awarded the maximum statutory damages of RMB 5 million (approx. US$ 700,000) compensation for economic loss, as well as reasonable expenses. Wu was held jointly and severally liable for the full amount, while Ling Company and Run Company were held jointly and severally liable for up to RMB 2.5 million (approx. US$ 350,000) and RMB 70,000 (approx. US$ 9,800), respectively.

On appeal, the Guangzhou Intellectual Property Court upheld the original judgment.  

Source: Yangcheng Pai

https://6nis.ycwb.com/app/template/displayTemplate/news/newsDetail/110059/54304061.html?isView=true

广州知识产权法院二审审结“蕉内”诉“蕉点”商标及网店装潢侵权案：“蕉点”因仿冒“蕉内”被顶格判赔500万元

日期：2026-09-23

三立人公司是“蕉内”系列注册商标的权利人，其官方旗舰店经长期运营形成了具有独特风格的整体营业形象。优某公司及吴某某申请注册“蕉点”“JIAODIAN蕉点”等商标，生产、销售与“蕉内”同类的商品；零某公司、润某公司亦通过网店宣传和销售相关产品。国家知识产权局自2023年11月起陆续宣告“蕉点”系列商标无效，相关裁定经行政诉讼终审维持，但优某公司仍继续申请近似商标并销售相关商品。三立人公司遂向广州市越秀区人民法院提起诉讼。

越秀法院一审认为，“蕉点”系列商标被宣告无效后，其注册商标专用权视为自始不存在，相关使用行为缺乏正当权利基础，构成商标侵权。被诉网店在背景配色、宣传标语、页面布局及商品展示等方面与“蕉内”官方旗舰店高度近似，并已造成消费者实际混淆，构成对有一定影响的网店整体装潢的仿冒，属于不正当竞争。

法院综合考虑优某公司系侵权商品的生产、销售源头，侵权持续时间长、涉及店铺多、销售规模大，以及同步仿冒品牌标识和网店装潢、主观攀附故意明显等因素，在法定赔偿幅度内顶格判令其赔偿经济损失及合理开支500万元；吴某某承担连带责任，零某公司和润某公司分别在250万元和7万元范围内承担连带责任。广州知识产权法院二审驳回优某公司等的上诉，维持原判。该判决现已生效。

资料来源：羊城派  2026-09-23

新闻链接：https://6nis.ycwb.com/app/template/displayTemplate/news/newsDetail/110059/54304061.html?isView=true