Artificial intelligence (AI) is transforming the basis of legal working. Modern AI models make case analysis faster and legal research more accessible, which is invaluable in a common law country such as India. Because AI models require much data for training, they often use existing databases. In the case of legal research, these include publicly available judgments and legislation as well as proprietary works that analyse such sources. This raises an important question of where AI fits within existing copyright law.
Senior Partner
Obhan & Associates
The recent judgment by the US District Court for the District of Delaware in the case of is the first of its kind to address the question of whether AI developers can train their AI models using copyrighted information, while relying on the exception of fair use. Although this judgment directly impacts the US copyright jurisprudence, it could have far-reaching persuasive influence on judicial precedents in other jurisdictions.
Thomson Reuters, the owner of Westlaw, accused Ross Intelligence, an AI-powered legal research firm, of violating its copyright. It was alleged that Ross failed to secure a licence from Westlaw to use its headnotes, which are the editorial summaries of case laws and are protected under copyright law. Ross obtained bulk memos through a third-party legal research company, LegalEase, and used the same to train its AI model. The bulk memos provided were created using Westlaw headnotes.
The court had to decide whether Westlaw possessed copyright on its headnotes and whether Ross had copied protectable elements of the copyrighted work, thereby infringing Westlaw’s copyright. The court was also asked to determine whether Ross’ use was covered by the fair use exception. The court dismissed Ross’s fair use argument, holding that the headnotes were sufficiently original to be protected under copyright law. Ross had infringed Westlaw’s copyright by using its headnotes to create similar content for training its own AI with the help of LegalEase.
Associate
Obhan & Associates
The court compared writing editorial summaries of a judgment to a sculptor shaping a block of marble – removing just enough to create something new and original. It was held that, although judicial opinions could not be protected under copyright law, the editorial discretion and commentary had sufficient creativity to be protected. It was also observed that the key number system used by Westlaw could be protected by copyright.
The court held that Thomson Reuters had sufficiently proved actual copying and substantial similarity, since Ross did use copyrighted work to create the content for training its AI. Although much of the case involved disputed facts and the drawing of inferences, which the court ordered to be left to a jury, the decisions in favour of the plaintiffs held that improving publicly available documents to produce curated legal content qualified for copyright protection and cannot be used without permission.
In dismissing the defendant’s claim of fair use, the court analysed the purpose and character of the use; the nature of the copyright work; the amount and importance of the work used, and the effect on the market value of the original work. The court held that Ross could not claim the defence of fair use because the use of the copyrighted material of Thomson Reuters by Ross was not transformative. It was also observed that the defendant’s AI tool was a direct competitor of Thomson Reuters’ Westlaw and therefore could be considered a market substitute.
In the past, courts have been lax when dealing with claims of fair use in areas such as software development, data analysis and research. However, this judgment establishes that the training of AI models by using substantial volumes of copyright material, especially when done by commercial competitors, will, in all likelihood, not satisfy the criteria for fair use.
It may be expected that this judgment will influence legal jurisprudence on the subject outside the borders of the United States. Courts in other jurisdictions may be influenced by the principles laid down in the case when they adjudicate AI-related copyright disputes. This may also lead to AI companies re-evaluating their approach to data licensing, especially if they use information from legal research platforms, such as Westlaw, LexisNexis, SCC Online and Manupatra, when they train their AI models.
Ashima Obhan is a senior partner and Aakanksha Singh is an associate at Obhan & Associates
Obhan & associatesAdvocates and patent agents
N – 94, second floor
Panchsheel park
New Delhi 110017, India
Contact details:
Ashima Obhan
T: +91 98 1104 3532
E: email@obhans.com ashima@obhans.com
