维基百科运营方维基媒体基金会在一宗涉及《纽约时报》、OpenAI和微软等公司的AI版权诉讼中提交法庭之友意见书，主张法院不要把人工智能训练和聊天机器人向用户复现内容混为一谈。基金会没有支持任何一方的整体立场，而是警告，过于简单化的裁决可能同时伤害开放知识与互联网内容生态。
维基媒体认为，网站抓取本身不应被视为版权侵权；未经许可将受版权保护的作品用于大型语言模型预训练，如果模型产出与原文“有所不同”，可以适用合理使用原则。但如果聊天机器人直接输出文章内容，取代读者访问原始作品，AI公司就应遵守授权要求。
这一主张与维基百科采用的知识共享署名—相同方式共享许可有关。维基媒体称，模型会复现以开放许可发布的维基百科文字，却未必遵守署名和相同方式共享等条件。基金会担心，若法院广泛认定AI训练一概不受限制，可能为忽视许可、削弱作者与读者之间内容流通开绿灯；反过来，若版权方获得过于宽泛的控制权，也可能阻碍对作品进行创新性、转化性使用。
维基媒体还指出，AI工具既带来风险，也能帮助维基百科编辑识别破坏性修改、版权违规和人肉信息，并找出需要改进的条目；但聊天机器人直接提供信息，也可能让用户不再访问维基百科及其他原始网站，削弱依赖社区维护的开放知识项目。
《纽约时报》2023年起诉OpenAI和微软，指控两家公司未经许可抓取大量新闻文章并用于训练AI系统；两家公司主张相关训练属于合理使用。维基媒体警告，法院若未能细致权衡训练与生成内容的差异，裁决可能影响未来AI开发、开放网络以及用户获取信息的方式。
本文由 AI宝库自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
Wikimedia Foundation, a Wikipedia operator, filed an amicus curiae opinion in an AI copyright suit involving companies such as The New York Times, Open AI and Microsoft, arguing that the court should not confuse artificial intelligence training and chat robots with reproducing to users. Instead of supporting the overall position of either party, the Foundation warned that simplistic decisions could harm open knowledge and Internet content ecology at the same time.
According to the Wiki media, website capture should not in itself be considered a copyright infringement; without permission, copyright-protected works are used for pre-training in large language models, and if model outputs are “different” to the original text, the principle of reasonable use can be applied. But if chat robots directly export the content of the article and replace readers with access to the original work, AI will comply with the authorization requirements.
This idea is related to the knowledge-sharing signature used in Wikipedia - the same way in which licences are shared. According to the Wikimedia, the model would be reproduced in open-licensed wikipedia of publish without necessarily complying with such conditions as signature and the same way of sharing. The Foundation is concerned that if the courts are widely satisfied that AI training is not restricted, it may open a green light to ignore licences and weaken the flow of content between authors and readers; in turn, if copyrighters are given too broad control, it may prevent the innovative and transformative use of the work.
Wikimedia also noted that the AI tool posed risks and helped Wikipedia editors to identify destructive modifications, copyright violations and human meat information and to identify entries that needed improvement; but chat robots provided information directly and could also discourage users from accessing Wikipedia and other original sites and weaken open knowledge projects that relied on community maintenance.
The New York Times filed charges against Open AI and Microsoft in 2023, accusing two companies of seizing a large number of news articles without permission and using them for training the AI system; the two companies argued that the training was a reasonable use. Wikimedia warned that court decisions that failed to carefully weigh the difference between training and production could affect future AI development, open networks and the way users receive information.
Auto-fetched and translated by AI Gem Box. All rights belong to the original author.