自几年前人工智能技术兴起以来，受版权保护的作品是否可以自由用于训练生成式人工智能模型一直是 数十起诉讼的核心问题。上周，美国地方法院宣布了两项关于Anthropic和Meta的判决，将合理使用抗 辩应用于人工智能训练，初步解答该问题的意见，然而这两个决定的矛盾性质说明，要真正解决这个问 题，还有很长的路要走。 尽管事实上这两起案件都是在同一个地区法院几天之内判决的，而且都裁定Meta和Anthropic的AI 训练 属于合理使用，但这些案件的判决截然不同，也并非像某些报道所称的那样，完全是AI公司的"胜利"。 带裂痕的共识：法官们"合理使用"认定的共同点 转换性使用"的模糊边界 两位法官都认定，根据第一项合理使用因素，将受版权保护的作品用于人工智能训练属于转化性使用。 他们还都得出结论，第一项合理使用因素对人工智能公司被告有利。 Anthropic案的Alsup法官基于以下三点考虑判定该使用属于转化性使用：（1）输出不侵权；（2）人工 智能训练与人类学习类似；（3）人工智能是改变游戏规则的技术（即具有"变革性"）。笔者认为， Alsup法官完全依赖这些因素是极其错误的。 序章：加州48小时内的"合理使用"双判 ...

Do reaction videos constitute fair use? That is the question at the heart of a new copyright infringement claim brought by Ethan Klein, an American YouTuber and podcaster https://t.co/GlNXLgNQBX ...

Core Viewpoint - AI companies, particularly Meta and Anthropic, have achieved significant legal victories regarding the use of copyrighted materials for AI training, but these rulings do not establish a blanket legality for such practices in the future [1][2][8]. Group 1: Legal Rulings - Meta won a motion for partial summary judgment in a case involving 13 authors who claimed copyright infringement due to the use of their books for training Llama AI models [1]. - Judge Vince Chhabria emphasized that the ruling does not imply that Meta's use of copyrighted materials is lawful, but rather that the plaintiffs failed to present compelling arguments [2]. - The rulings are significant as they are among the first to provide substantive legal analyses on the fair use doctrine in the context of AI [8]. Group 2: Fair Use Doctrine - The core issue revolves around whether AI companies' use of protected content qualifies as fair use, which allows for certain uses of copyrighted work without permission [2]. - The fair use evaluation considers four key factors, with Meta's ruling focusing on the impact of AI on the existing publishing market [2][5]. - Judge Chhabria noted that while AI-generated works could diminish the market for human-created books, the plaintiffs did not provide sufficient evidence of harm [6]. Group 3: Industry Implications - The victories for AI companies may reduce the need for costly licensing agreements with content creators, which has raised concerns among authors [3]. - A group of authors has publicly urged publishers to take a stronger stance against AI, highlighting the lack of permission and compensation for the use of their works [4]. - The rulings may influence future cases, as they set precedents that judges can reference in similar copyright disputes involving AI [9][10].

It's a fascinating read. And you go into the intricacies of how entropic, first of all, did it. And they did it perhaps with a mixture of pirated book piece.But then they actually started buying physical books, taking out the spine and copying them into the computer. Yes. So they went out and they they acquired pirated copies of more than 7 million books to try in their models.After a short while, they thought, well, maybe there's a better way to do this. And they bought the physical used copies from distri ...

Mark Zuckerberg’s Meta has won the backing of a judge in a copyright lawsuit brought by a group of authors, in the second legal victory for the US artificial intelligence industry this week.The writers, who included Sarah Silverman and Ta-Nehisi Coates, had argued that the Facebook owner had breached copyright law by using their books without permission to train its AI system.The ruling follows a decision on Monday that Anthropic, another major player in the AI field, had not infringed authors’ copyright.Th ...

Core Viewpoint - A federal judge ruled in favor of Meta in a lawsuit regarding the training of AI models on copyrighted works, determining that such use falls under the "fair use" doctrine of copyright law [2][3]. Group 1: Legal Rulings - Judge Vince Chhabria issued a summary judgment, allowing Meta to avoid a jury trial, and concluded that the training of AI models on copyrighted books was legal under fair use [2]. - The ruling is part of a broader trend favoring the tech industry, as seen in a similar case involving Anthropic, although both rulings are limited in scope [3]. - Judge Chhabria emphasized that this decision does not imply that all AI training on copyrighted materials is lawful, but rather that the plaintiffs failed to present adequate arguments and evidence [4][5]. Group 2: Market Impact and Evidence - The judge noted that the plaintiffs did not provide meaningful evidence to demonstrate that Meta's actions harmed the market for the authors' works, which is crucial in copyright cases [8]. - The ruling highlighted that the transformative nature of Meta's AI models, which do not simply reproduce the authors' books, played a significant role in the decision [5]. Group 3: Industry Context - The outcomes of the cases involving Meta and Anthropic are part of ongoing legal battles faced by technology companies regarding the use of copyrighted materials for AI training, with other lawsuits pending against companies like OpenAI and Midjourney [9]. - Judge Chhabria pointed out that fair use defenses are highly case-specific, suggesting that different industries may have varying strengths in their fair use arguments [10].