联邦法官William Alsup在2026年5月对Bartz诉Anthropic案作出裁定，认定AI公司Anthropic在训练其Claude大语言模型时，通过盗版手段非法获取并使用了包括多位斯坦福大学教授在内的作者作品1。尽管法官判定合法获取的书籍用于AI训练属于“合理使用”（fair use），但明确禁止通过盗版建立永久中央图书馆的行为1。这一裁决涉及近50万件注册作品的作者和出版商，此前达成的15亿美元和解方案虽被Authors Alliance称为涉及AI公司的最大版权和解案，但遭到部分创作者反对1。
针对和解金额不足的问题，Tobias Wolff等28位作者选择退出集体诉讼，并于同月提起新的非集体诉讼Cruz v. Anthropic1。原告指控Anthropic从LibGen和PiLiMi等影子图书馆非法下载、复制及分发受版权保护的书籍1。新诉讼寻求每件侵权作品最高15万美元的法定赔偿，这一数额远高于Bartz和解中约3000美元的补偿标准1。对此，学者Paul Goldstein指出解决AI版权纠纷的核心难点在于高昂的交易成本，而Authors Guild正与国会合作推动要求许可的国家立法进程1。
A federal judge has determined that AI company Anthropic illegally acquired and utilized works by several Stanford University professors to train its Claude model, a ruling stemming from the May 2026 case Bartz v. Anthropic 1. While Judge William Alsup concluded that using legally obtained books for large language model training constitutes fair use, he explicitly prohibited the creation of permanent central libraries through piracy 1. The plaintiffs alleged that Anthropic downloaded, copied, and distributed copyrighted books from shadow libraries such as LibGen and PiLiMi 1.
The Bartz case resulted in a settlement valued at $1.5 billion, covering authors and publishers of nearly 500,000 registered works, which Authors Alliance described as the largest copyright settlement involving an AI company 1. However, some authors deemed this amount insufficient, noting that it equated to approximately $3,000 per work 1. Consequently, Tobias Wolff and 27 other authors opted out of the settlement and filed a new non-class action lawsuit, Cruz v. Anthropic, in May 2026, seeking statutory damages of up to $150,000 for each infringed work 1.
Expert commentary highlights the structural challenges in resolving AI copyright disputes. Paul Goldstein noted that transaction costs remain the core difficulty in settling these issues, while the Authors Guild is currently collaborating with Congress to draft national legislation mandating licensing 1.
评论
还没有评论，欢迎留下第一条。