[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$fIFyva02BP1ZiyaQPUfcIlyqX6kgchuS5fGbuKuEQqQ8":3},{"code":4,"msg":5,"data":6},200,"操作成功",{"id":7,"title":8,"content":9,"digest":10,"source":10,"coverPath":11,"thumbsCoverPath":12,"isTop":13,"isShow":14,"baseClick":13,"clickCount":15,"createTime":16,"typeId":17,"isNewest":18,"newsInfoTypeRespVo":19,"voiceUrl":22,"voiceSize":23,"taskId":24,"releaseTime":25,"titleEn":26,"contentEn":27,"voiceUrlEn":28,"taskIdEn":29,"voiceSizeEn":30},1232,"合理使用非万能数据授权机制成关键","\u003Cp>\u003Cstrong style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">在人工智能技术与版权法交汇的关键节点，Meta公司在一场由图书作者发起的盗版诉讼中赢得部分合理使用抗辩。美国加州北区地方法院法官文斯·查布里亚虽支持Meta公司关于合理使用的核心主张，但明确指出该裁决仅适用于本案特定情形，更通过详细法律分析为未来人工智能开发者面临的版权挑战划定关键路径。这场涉及作家理查德·卡德雷、莎拉·西尔弗曼和克里斯托弗·戈登等13位创作者的集体诉讼，因Meta被指控使用盗版书库训练其Llama模型而引发行业高度关注。\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"color: rgb(255, 153, 0); font-size: 18px;\" class=\"ql-lineHeight-1-75\">焦点：AI训练行为是否构成“合理使用”\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">自2023年起，全球范围内掀起针对AI企业的版权诉讼浪潮。该案原告指控Meta通过BitTorrent网络获取数百万本盗版电子书，将其作为训练Llama大型语言模型的核心数据源。Meta虽承认使用非官方渠道获取的文本资料，但坚决否认直接侵权，并以“合理使用”原则作为核心抗辩理由。双方于2025年3月分别提交简易裁决动议，焦点集中于AI训练行为是否构成版权法意义上的“合理使用”。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">法院法官文斯·查布里亚在裁决中系统分析了美国版权法第107条规定的合理使用四要素：\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">使用目的与性质：认定Meta对书籍的使用具有“转化性”，因其通过算法提取语言模式而非复制内容，最终产物LLM模型属于“创新工具”，功能远超出原始文本的文学表达。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">作品性质：承认涉案书籍多为虚构类文学创作，但未过度强调该因素对合理使用的影响。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">使用量：指出Meta虽系统性复制全书内容，但具体使用场景中仅提取统计特征而非实质性文本片段。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">市场损害：判定原告未能证明实际或潜在市场损害，成为案件转折关键。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">值得注意的是，该案的原告提出的三项市场损害主张均遭到法院驳回：直接替代风险：通过专家测试证实，即便在“对抗性提示”场景下，Llama模型也无法从单本受保护作品中生成超过50个连续单词，不具备实质性替代能力；许可市场损害：法院明确拒绝将“假设存在的AI训练许可市场”纳入赔偿考量，强调版权法仅保护现实存在的市场，而非潜在商业机会；市场稀释理论：法官虽承认该理论可能构成合理使用挑战，但指出原告仅提出概念而未提供实证数据，例如AI生成内容对图书销量的具体影响统计。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"color: rgb(255, 153, 0); font-size: 18px;\" class=\"ql-lineHeight-1-75\"> 意义：为人工智能开发者敲响警钟\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">尽管Meta公司明确胜诉，但法院的裁决对其来说却是一场苦乐参半的胜利。该动议只涵盖了版权主张的一部分，因为Meta公司涉嫌分发盗版书籍的行为并未被纳入审理范围。此外，该裁决只适用于本案中的13位署名作者。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">此次“胜诉”并不意味着合理使用抗辩在其他人工智能培训版权案件中也能成立。事实上，法院暗示，Meta公司和其他公司可能需要接受为此类用途获取内容授权的理念。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Meta公司认为，负面的合理使用裁决可能会阻止人工智能技术的发展，因为人工智能模型需要海量的数据进行训练。然而，法院驳回了这一推论。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">查布里亚在裁决书中写道：“不利的版权裁决会让这项技术停滞不前的说法是荒谬的。这些产品预计将为其开发公司带来数十亿甚至数万亿美元的收益。如果使用受版权保护的作品来训练模型真像这些公司所说的那样有必要，他们理应设法补偿版权持有人。”\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">法院作出有利于Meta公司的裁决，更多是因为作者们未能充分论证市场稀释的主张，而不是人工智能培训的合理使用明显获胜。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">查布里亚表示，基于现有庭审记录，法院别无选择，只能作出有利于Meta公司的简易裁决。不过，这当然并不意味着类似的论点在其他案件中也会成立。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">裁决书明确指出：“现在应当明确，这一裁决并不代表Meta公司使用受版权保护的材料来训练其语言模型的行为是合法的。它只代表这些原告提出了错误的论点，并且未能提供支持正确论点的证据。”\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">随着全球AI监管趋严，该案凸显技术发展与版权保护间的深层矛盾。Meta虽暂时避免直接侵权责任，但法院对授权机制的暗示，预示着行业或将迎来数据使用规范化的转折点。如何在创新需求与权益保障间寻求平衡，仍将是法律与技术领域持续探索的命题。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"color: rgb(187, 187, 187);\" class=\"ql-lineHeight-1-75\">【新闻来源】中国新闻出版广电报（中国保护知识产权网） \u003C\u002Fspan>\u003Ca href=\"https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002F%E5%90%88%E7%90%86%E4%BD%BF%E7%94%A8%E9%9D%9E%E4%B8%87%E8%83%BD%E6%95%B0%E6%8D%AE%E6%8E%88%E6%9D%83%E6%9C%BA%E5%88%B6%E6%88%90%E5%85%B3%E9%94%AE\u002Far-AA1KuieB?ocid=msedgntphdr&cvid=0528260646f44b889f8a9ae6e76d3e66&ei=20\" rel=\"noopener noreferrer\" target=\"_blank\" style=\"color: rgb(187, 187, 187);\" class=\"ql-lineHeight-1-75\">http:\u002F\u002Fu5a.cn\u002FF8YCN\u003C\u002Fa>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(187, 187, 187);\" class=\"ql-lineHeight-1-75\">（本网转发此文章，旨在为读者提供更多的信息资讯，所涉内容不构成投资、消费建议。文章事实如有疑问，请与有关方核实，文章观点非本网观点，仅供读者参考。）\u003C\u002Fspan>\u003C\u002Fp>","","https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F08\u002F517b74833b7848d7b55e0cc191cc3791\u002FAI领域.jpg","https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F08\u002Fthumbs\u002F517b74833b7848d7b55e0cc191cc3791\u002FAI领域.jpg",0,1,225,"2025-08-15 17:44",2,false,{"id":17,"name":20,"enName":21},"芯位视野","Xinwei Vision","https:\u002F\u002Fxinwei-dev-test.oss-cn-shenzhen.aliyuncs.com\u002Fintelligent\u002Faudio%3Aab6edb74-d665-4b6d-91de-39bf1dd06c0d%3A0.wav?Expires=1755255082&OSSAccessKeyId=LTAI5tNvY2RkKjZw4LLWsrPK&Signature=sAxqDRc1ccgHRYHcr5yazh%2FMTWM%3D",8625360,"ab6edb74-d665-4b6d-91de-39bf1dd06c0d","2025-08-15 17:38","Reasonable use of non-universal data authorization mechanisms becomes key.","\u003Cp>\u003Cstrong style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">At the critical junction where artificial intelligence technology and copyright law intersect, Meta won part of its fair use defense in a copyright infringement lawsuit initiated by book authors. Judge Vince Chhabria of the U.S. District Court for the Northern District of California supported Meta's core argument regarding fair use, but clearly stated that the ruling applies only to the specific circumstances of this case. Through detailed legal analysis, the court also outlined a key path for future challenges faced by AI developers regarding copyright issues. This class-action lawsuit involving 13 creators, including author Richard Kadrey, Sarah Silverman, and Christopher Gordon, has drawn significant industry attention due to Meta being accused of using a pirated book library to train its Llama model.\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"color: rgb(255, 153, 0); font-size: 18px;\" class=\"ql-lineHeight-1-75\">Focus: Whether AI training behavior constitutes \"fair use\"\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Since 2023, a wave of copyright lawsuits against AI companies has swept across the globe. The plaintiffs in this case accused Meta of obtaining millions of pirated e-books through the BitTorrent network, using them as the core data source for training the Llama large language model. Although Meta admitted using text materials obtained through non-official channels, it strongly denied direct infringement and used the \"fair use\" principle as its main defense. Both parties submitted motions for summary judgment in March 2025, with the focus on whether AI training behavior constitutes \"fair use\" under copyright law.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Judge Vince Chhabria analyzed the four factors of fair use under U.S. Copyright Act Section 107 in his ruling:\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Purpose and nature of use: The court determined that Meta's use of the books was \"transformative,\" as it extracted language patterns through algorithms rather than copying content, and the final product, the LLM model, is an \"innovative tool\" with functions far beyond the literary expression of the original text.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Nature of the work: The court acknowledged that the works in question were mostly fictional literary creations, but did not overemphasize the impact of this factor on fair use.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Amount used: The court pointed out that although Meta systematically copied the entire content of the books, it only extracted statistical features rather than substantial text fragments in specific usage scenarios.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Market harm: The court ruled that the plaintiffs failed to prove actual or potential market harm, which became a turning point in the case.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Notably, all three market harm claims raised by the plaintiffs were rejected by the court: Direct substitution risk: Expert testing confirmed that even in \"adversarial prompt\" scenarios, the Llama model could not generate more than 50 consecutive words from a single protected work, indicating no substantial substitution capability; License market harm: The court explicitly refused to consider the \"hypothetical AI training license market\" in compensation, emphasizing that copyright law protects only existing markets, not potential commercial opportunities; Market dilution theory: While the judge acknowledged this theory might pose a challenge to fair use, he noted that the plaintiffs presented only concepts without providing empirical data, such as specific statistics on the impact of AI-generated content on book sales.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"color: rgb(255, 153, 0); font-size: 18px;\" class=\"ql-lineHeight-1-75\"> Significance: A warning bell for AI developers\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Although Meta won the case, the court's ruling was a bittersweet victory for the company. This motion covered only part of the copyright claims, as Meta's alleged distribution of pirated books was not included in the scope of review. Additionally, the ruling applies only to the 13 named authors in this case.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">This \"victory\" does not mean that the fair use defense would hold in other AI training copyright cases. In fact, the court implied that Meta and other companies may need to accept the idea of obtaining content licenses for such purposes.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Meta argued that negative fair use rulings could hinder the development of AI technology, as AI models require massive amounts of data for training. However, the court rejected this inference.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Chhabria wrote in his ruling: \"The claim that adverse copyright rulings would stall this technology is absurd. These products are expected to bring billions or even trillions of dollars in revenue for their developers. If using copyrighted works to train models is truly necessary as these companies claim, they should have found a way to compensate copyright holders.\"\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">The court's ruling in favor of Meta was more because the authors failed to adequately argue the market dilution claim, rather than because the fair use of AI training was clearly established.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Chhabria said that based on the current trial record, the court had no choice but to issue a summary judgment in favor of Meta. However, this certainly does not mean that similar arguments will hold in other cases.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">The ruling clearly states: \"It should now be clear that this ruling does not mean that Meta's use of copyrighted material to train its language model is legal. It only indicates that these plaintiffs made incorrect arguments and failed to provide evidence supporting the correct arguments.\"\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">As global AI regulation tightens, this case highlights the deep contradictions between technological development and copyright protection. Although Meta temporarily avoided direct infringement liability, the court's implication about the authorization mechanism suggests that the industry may be approaching a turning point toward data usage standardization. How to balance innovation needs and rights protection remains a continuous exploration in the fields of law and technology.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"color: rgb(187, 187, 187);\" class=\"ql-lineHeight-1-75\">【News Source】China News Publishing & Broadcasting Daily (China Intellectual Property Protection Network) \u003C\u002Fspan>\u003Ca href=\"https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002F%E5%90%88%E7%90%86%E4%BD%BF%E7%94%A8%E9%9D%9E%E4%B8%87%E8%83%BD%E6%95%B0%E6%8D%AE%E6%8E%88%E6%9D%80%E6%9C%BA%E5%88%B6%E6%88%90%E5%85%B3%E9%94%AE\u002Far-AA1KuieB?ocid=msedgntphdr&cvid=0528260646f44b889f8a9ae6e76d3e66&ei=20\" rel=\"noopener noreferrer\" target=\"_blank\" style=\"color: rgb(187, 187, 187);\" class=\"ql-lineHeight-1-75\">http:\u002F\u002Fu5a.cn\u002FF8YCN\u003C\u002Fa>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(187, 187, 187);\" class=\"ql-lineHeight-1-75\">（This article is reposted by this site to provide readers with more information. The content does not constitute investment or consumption advice. If there are any questions about the facts of the article, please verify with the relevant parties. The views expressed in the article are not the views of this site and are for reference only.）\u003C\u002Fspan>\u003C\u002Fp>","https:\u002F\u002Fxinwei-dev-test.oss-cn-shenzhen.aliyuncs.com\u002Fintelligent\u002Faudio%3A93da9d1d-7da9-43a7-a3eb-338b0917b352%3A0.wav?Expires=1774838494&OSSAccessKeyId=LTAI5tNvY2RkKjZw4LLWsrPK&Signature=qLrGUGmzbkqWUtYmeyL%2BjQS6fXg%3D","93da9d1d-7da9-43a7-a3eb-338b0917b352",10339264]