US Judge Approves $1.5B Anthropic Copyright Settlement

01The Story新闻事实

A US federal judge approved a $1.5 billion settlement resolving class-action claims that Anthropic used pirated books to train its Claude chatbot. Anthropic will pay authors about $3,000 per book, covering over 482,000 titles.美国联邦法官批准15亿美元和解协议，解决Anthropic使用盗版图书训练Claude机器人的诉讼。公司将向作者按每本约3000美元支付赔偿，��盖超48.2万本图书。

US District Judge Araceli Martínez-Olguín granted final approval to the $1.5B settlement on July 20, 2026.美国联邦地��法官阿拉塞莉·马丁内斯-奥尔金于2026年7月20日批准15亿美元和解案。
Anthropic will compensate affected authors and publishers approximately $3,000 per pirated book used for training.Anthropic将向受影响作者及出版商按每本盗版图书约3000美元支付赔偿。
Over 91% of the 482,000 eligible books covered under the ruling have been claimed by rightsholders.裁决涵盖的48.2万本图书中，超过91%已被权利人申领赔偿。
Plaintiff attorney Justin Nelson called the result the largest known copyright recovery in history.原告律师贾斯汀·内尔森称该结果为历史上最大的版权赔偿追偿案。

The $1.5B settlement penalizes illegal dataset acquisition while preserving the legal defense that training AI on legally obtained content constitutes fair use.15亿美元和解惩罚了非法盗版数据集获取，同时保留了在合法内容上训练AI属于合理使用的法律抗辩。

025 Dimensions维度分析

Cause因果关系

Dataset Origin vs. Training Fair Use数据集来源与训练合理使用

Courts distinguished between training AI models on lawful purchases and acquiring pirated texts from shadow libraries.法院对在合法购买文本上训练AI模型与从盗版网站获取图书作出了区别认定。

Pirated Books Covered涵盖盗版图书数量

482,000 titles48.2万本

2024-20262024-2026年

Retired Judge William Alsup ruled that training AI models on lawfully purchased books constitutes fair use.已退休的阿尔苏普法官曾裁定在合法购买的图书上训练AI模型属于合理使用。
Liability stemmed specifically from Anthropic acquiring millions of copyrighted works through unauthorized pirate websites.法律责任专门源于Anthropic通过未经授权的盗版网站获取数百万本受版权保护的作品。

Stakeholder impact利益相关方

Author Payouts and High Claim Rate作者赔偿标准与高申领率

Direct cash payments provide unprecedented individual compensation for creators whose works were scraped for AI training.直接现金赔偿为作品被抓取用于AI训练的创作者提供了创纪录的个人补偿。

Per-Book Compensation单本图书赔偿标准

20262026年3,000 USD

2026-072026年7月91 percent

Affected rightsholders will receive approximately $3,000 per book included in Anthropic's training repositories.受影响的权利人将就包含在Anthropic训练库中的每本图书获得约3000美元赔偿。
A 91% participation rate demonstrates broad alignment and rapid distribution across authors and publishing houses.91%的申领率体现了作者与出版商之间的高参与度及快速分发进展。

Comparison对比

Record Financial Scale of Copyright Recovery创纪录的版权追偿金额规模

The $1.5 billion settlement marks the largest single copyright class-action recovery recorded in technology history.15亿美元的和解创下了科技与版权史上最大的单笔集体诉讼赔偿金额。

Total Settlement Amount和解金总额

2026-072026年7月1.5B USD

Plaintiff counsel confirmed the recovery surpasses previous copyright class-action payouts in the digital era.原告律师证实，该赔偿金超越了数字时代以往所有的版权集体诉讼和解金额。
The class action was originally initiated in 2024 by authors Andrea Bartz, Kirk Wallace Johnson, and Charles Graeber.该集体诉讼最初由作家安德里亚·巴茨、柯克·华莱士·约翰逊及查尔斯·格雷伯于2024年发起。

Contradiction矛盾之处

Contrast with Ongoing AI Copyright Cases与其他在审AI版权诉讼的对比

While Anthropic resolved its pirated dataset dispute, active litigation against OpenAI and Microsoft leaves core legal doctrines open.尽管Anthropic解决了盗版数据集纠纷，但针对OpenAI和微软的诉讼仍未解决核心法律问题。

Major lawsuits by the Authors Guild and The New York Times against OpenAI remain active in federal courts.美国作家协会及《纽约时报》针对OpenAI的大型诉讼仍在美国联邦法院审理中。
Because this resolution is a private settlement, it creates no binding legal precedent for other pending court cases.由于本次和解属于私下协议，因此不会对其他在审法院案件构成约束性判例。

Future signal前瞻信号

Mandatory Erasure and Licensing Shift强制数据删除与商业授权转型

Settlement obligations require deleting pirated repositories, driving AI developers toward transparent commercial licensing.和解条款要求彻底删除盗版数据库，促使AI开发商向透明的商业授权合作转型。

Pirated Repositories Purged盗版数据库清理比例

100 percent100%

20262026年

The settlement mandates that Anthropic permanently delete all downloaded pirated book repositories from its storage.和解协议强制要求Anthropic从其存储系统中永久删除所有下载的盗版图书数据库。
AI developers are expanding direct commercial licensing agreements with rightsholders to reduce dataset liability.AI开发商正在扩大与权利人的直接商业授权协议，以降低数据集法律风险。

03Timeline时间线

2024-08 Authors file class-action suit against Anthropic in US court.作者在美国法院对Anthropic提起集体诉讼。
2025-06 Court rules training AI on lawfully purchased books is fair use.法院裁定在合法购买的图书上训练AI属于合理使用。
2025-09-25 US judge grants preliminary approval to $1.5B settlement.美国法官初步批准15亿美元和解协议。
2026-07-17 Anthropic counsel calls fair use ruling a key landmark.Anthropic法务官称合理使用裁决具有里程碑意义。
2026-07-20 Judge Martínez-Olguín grants final approval to $1.5B payout.马丁内斯-奥尔金法官最终批准15亿美元和解案。

Key takeaway核心结论

Anthropic's $1.5B settlement proves pirating training data creates severe liability, even as training on legal data remains protected fair use.Anthropic达成的15亿美元和解表明，盗取训练数据将带来沉重法律责任，即便合法训练仍受合理使用保护。

04What to Watch后续关注

Rulings in active copyright class actions against OpenAI and Microsoft.针对OpenAI与微软的在审版权集体诉讼裁决。
Distribution timeline of $1.5 billion compensation to eligible authors.向合规作者发放15亿美元赔偿金的具体进度。
Adoption of formal commercial licensing deals between AI labs and publishers.AI实验室与出版商之间商业授权协议的签订。

05Uncertainties不确定因素

Whether non-settling AI firms can establish total fair use defenses in court.未和解AI公司能否在法庭上确立完整合理使用抗辩。
Potential future liability claims regarding AI model generated output similarity.关于AI生成内容相似度的潜在后续侵权诉讼风险。

06Sources信息来源

Share this analysis 分享此深度分析

X Facebook LinkedIn WhatsApp