从法庭到画布：人工智能版权案如何影响首场大型人工智能艺术品拍卖会
来源：广东中策知识产权研究院 发布日期：2025-03-03 阅读：11次
On Tuesday, February 11, the U.S. Court of Appeals for the Third Circuit Judge Stephanos Bibas issued a partial summary judgment in Thomson Reuters v. Ross Intelligence, finding that Ross Intelligence (Ross) infringed on Thomson Reuters’ copyrighted case law headnotes and rejecting Ross’ fair use defense. The court found that Ross infringed Thomson Reuters’ copyrights in 2,243 Westlaw case law headnotes by using them to train its non-generative artificial intelligence (AI) legal research tool, determining that these headnotes met the copyright threshold of originality. Judge Bibas likened Westlaw’s editorial process to that of a sculptor carving away excess material from a marble slab, underscoring that the selection and phrasing of headnotes constitute protected expression.
2 月 11 日星期二，美国第三巡回上诉法院法官 Stephanos Bibas 在汤森路透诉 Ross Intelligence 一案中做出了部分简易判决，认定 Ross Intelligence（Ross）侵犯了汤森路透的判例法标题注释版权，并驳回了 Ross 的合理使用抗辩。法院认定 Ross 侵犯了汤森路透对 2,243 篇 Westlaw 判例法注释的版权，因为 Ross 将这些注释用于训练其非生成式人工智能法律研究工具，并认定这些注释达到了原创性的版权门槛。比巴斯法官将Westlaw的编辑过程比作雕刻师从大理石板上雕刻掉多余的材料，强调标题注释的选择和措辞构成受保护的表达。
Although two of the four fair use factors favored Ross, the court placed significant weight on the first and fourth factors: the purpose of Ross’ use was commercial and directly competed with Westlaw, undermining its market. Originally filed in 2020, this case was one of the first to address the legality of AI tools trained on copyrighted material obtained without permission or licensing.
尽管四个合理使用因素中有两个有利于Ross，但法院对第一个和第四个因素给予了极大的重视：Ross的使用目的是商业性的，直接与 Westlaw 竞争，破坏了其市场。该案最初于 2020 年提起诉讼，是最早涉及未经许可或授权而在受版权保护的材料上训练人工智能工具的合法性问题的案件之一。
This decision signals the growing judicial scrutiny of generative AI platforms and countless allegations that the AI models were trained on copyrighted material without permission. Many expect future fair use defenses to mirror Ross’ arguments, particularly regarding the extent to which AI-generated outputs compete with the copyright holder’s market, as identified in factor four of the fair use test. Notably, programs like DALL-E, Stable Diffusion, and Midjourney have drawn similar allegations and lawsuits from artists who argue that AI companies trained their models on their copyrighted works without permission or compensation.
这一判决预示着对生成式人工智能平台的司法审查日益严格，以及对人工智能模型未经许可在受版权保护的材料上进行训练的无数指控。许多人预计，未来的合理使用抗辩将反映Ross的论点，特别是关于人工智能生成的产出与版权持有者市场竞争的程度，正如合理使用测试要素四所确定的那样。值得注意的是，像《DALL-E》、《Stable Diffusion》和《Midjourney》这样的程序也招致了艺术家们类似的指控和诉讼，他们认为人工智能公司在未经许可或补偿的情况下用他们的版权作品训练模型。
AI in the Art World
艺术界的人工智能
The controversy surrounding AI and copyright extends beyond the courtroom to the art market, where AI-generated works are increasingly contested. Christie’s is facing backlash over its upcoming AI art auction, Augmented Intelligence, set to take place in New York. An online open letter, signed by over 5,000 individuals and counting, argues that some generative artworks in the sale were created using programs trained on copyrighted works that exploit human artists’ labor.
围绕人工智能和版权的争议已从法庭延伸到艺术市场，在艺术市场上，人工智能生成的作品受到越来越多的争议。佳士得拍卖行即将在纽约举行的人工智能艺术品拍卖会--“增强智能”（Augmented Intelligence）正面临反弹。一封由 5000 多人签名的在线公开信认为，拍卖会上的一些生成艺术品是使用在受版权保护的作品上训练的程序创作的，剥削了人类艺术家的劳动。
The sale, open for bids from February 20 to March 5, will feature over 20 lots of AI-generated artworks across various mediums, including sculptures, paintings, and prints, some dating back to the 1960s. Among the works included is Untitled Robot Painting by Alexander Reben, an artist who completed the first OpenAI artist residency last year. Reben’s work integrates live performance art with generative AI. During the auction, a 12-foot-tall robot will create an oil painting on a 10×12-foot acrylic canvas in real time at Christie’s New York location. The robot’s painting process will be guided by Reben’s AI model, with each new bid—starting at $100—triggering the robot to paint a blank section of the canvas. The final price, according to Christies, may reach $1.7 million.
此次拍卖会于2月20日至3月5日开放竞拍，将展出20多件人工智能生成的艺术品，包括雕塑、绘画和版画等各种媒介，其中一些可以追溯到20世纪60年代。其中包括亚历山大-雷本（Alexander Reben）的《无题机器人绘画》（Untitled Robot Painting）。雷本的作品将现场表演艺术与人工智能生成技术相结合。拍卖期间，一个 12 英尺高的机器人将在佳士得纽约分部的 10×12 英尺丙烯酸画布上实时创作一幅油画。机器人的作画过程将由 Reben 的人工智能模型指导，每一次新的出价（起价 100 美元）都会触发机器人在画布的空白处作画。据克里斯蒂拍卖行称，最终价格可能会达到 170 万美元。
Despite mounting criticism, the auction is planned to proceed at the time of this writing. Christie’s anticipates a total revenue around $600,000 and has announced that it will accept cryptocurrency as payment for most of the works.
尽管批评声音越来越多，但在撰写本文时，拍卖仍计划继续进行。佳士得拍卖行预计总收入约为 60 万美元，并宣布将接受加密货币作为大部分作品的付款方式。
Brush Up and Adapt
改进与调整
The ruling in Thomson Reuters v. Ross Intelligence adds legal weight to concerns about AI’s impact on creative and commercial markets, including the sale of fine art through auction houses and galleries, and its potential to compete directly with copyright holders. If courts continue to rule against unlicensed AI training on copyrighted materials, auction houses, galleries, collectors, and artists utilizing AI may face greater scrutiny over the provenance of AI-generated artworks, along with reputational risks. Whether AI-generated content and artworks can be framed as original and transformative remains uncertain. In light of this decision, both artists and AI companies may need to brush up on and adapt their approaches to improve transparency in data sourcing and training while ensuring copyright compliance.
汤姆森-路透社诉罗斯智能公司案的裁决为关于人工智能对创意和商业市场的影响，特别是通过拍卖行和画廊出售艺术品以及其可能直接与版权持有者竞争的担忧，增加了法律权重。如果法院继续裁定未经授权的人工智能训练使用了受版权保护的材料，拍卖行、画廊、收藏家和使用人工智能的艺术家可能会面临对人工智能生成艺术品来源的更大审查，以及声誉风险。人工智能生成的内容和艺术作品是否能够被界定为原创和具有变革性仍然不确定。鉴于这一裁决，艺术家和人工智能公司可能需要改进和调整他们的做法，以改进数据来源和训练的透明度，同时确保遵守版权法规。