Introduction
The legal landscape surrounding the use of copyrighted material in training generative AI tools is evolving, as demonstrated by a recent ruling in favor of OpenAI. The case [2] [3] [4] [5] [6] [7], involving allegations of Digital Millennium Copyright Act (DMCA) violations [3], highlights the complexities of copyright law in the context of AI development and the challenges plaintiffs face in establishing concrete injury.
Description
On November 7, 2024 [1] [3], the US District Court for the Southern District of New York ruled in favor of OpenAI [3], dismissing a copyright lawsuit filed by Raw Story Media [2] [6], Inc. and AlterNet Media Inc [1] [2] [3] [6] [7]. The plaintiffs alleged that OpenAI’s use of their original journalistic content to train the language model powering ChatGPT violated the Digital Millennium Copyright Act (DMCA), specifically claiming that essential copyright management information (CMI) [3], such as author names and titles [3], was removed from their works. The court emphasized that Section 1202(b) of the DMCA does not grant copyright owners control over future versions of their works, allowing OpenAI to reproduce or create derivative works without liability as long as the CMI remained intact.
Presided over by Judge Colleen McMahon, the court highlighted the necessity of demonstrating a concrete injury-in-fact for Article III standing [7]. Although the plaintiffs contended that their injury was closely related to copyright infringement [7], the court disagreed [2] [6] [7], stating that the DMCA primarily protects the integrity of CMI rather than the use of the work itself [7]. The plaintiffs failed to provide evidence of actual harm or adverse effects resulting from the alleged DMCA violation [7], particularly regarding monetary damages or the risk of future harm [3]. The court noted the minimal likelihood of ChatGPT generating plagiarized content from their articles and indicated that the plaintiffs did not meet the double-scienter knowledge standard required under Section 1202(b)(1), which necessitates awareness of the unauthorized removal of CMI and its potential to lead to infringement.
Additionally, the court rejected the plaintiffs’ claim for injunctive relief based on the potential risk of future harm [7], asserting that any risk must be imminent and substantial [5] [7]. The court concluded that the real injury sought by the plaintiffs was the unauthorized use of their articles for training ChatGPT [7], which did not fall under the DMCA’s protections [7]. The judge also suggested that the claims were based on an outdated model of ChatGPT [1], indicating that a revised complaint referencing the current model and additional evidence might be more persuasive [1].
While the plaintiffs were granted leave to amend and refile their complaint [7], the judge expressed skepticism about their ability to establish a cognizable injury [7]. This case reflects a growing trend of litigation concerning the use of copyrighted material in training generative AI tools [2] [6], with similar cases [1] [2] [6], such as one filed by The New York Times against Microsoft and OpenAI [2] [6], also pending in the same jurisdiction [2] [6].
In contrast [5] [7], another case [3] [7], The Intercept Media [5] [7], Inc. v. OpenAI [1] [2] [3] [4] [5] [6] [7], allowed Section 1202(b)(1) claims to proceed [5] [7], suggesting that the presence of specific examples of outputs generated by ChatGPT may have satisfied the standing requirement [7]. This highlights differing judicial approaches to DMCA claims in the context of AI training [5]. In a separate case [7], Andersen v [5] [7]. Stability AI [5] [7], a California court dismissed DMCA allegations [5] [7], requiring identicality between the original work and the copy for a claim to succeed [7]. This issue is currently under appeal [5] [7], which may influence future DMCA claims in AI training data cases [7].
The overall trend indicates that if courts adopt a stringent interpretation of standing requirements [7], DMCA claims may become less frequent unless there is clear evidence of infringing outputs generated without CMI [7]. The Supreme Court’s ruling in TransUnion v [5] [7]. Ramirez [5] [7], which established a concrete injury requirement [5] [7], could further impact cases involving the Copyright Act and related statutory claims [7]. OpenAI was also granted an extension to respond to the plaintiffs’ request [4], indicating ongoing developments in this legal landscape.
Conclusion
The ruling in favor of OpenAI underscores the challenges plaintiffs face in proving concrete injury under the DMCA in the context of AI training. As courts continue to interpret standing requirements stringently, the frequency of such claims may decrease unless clear evidence of infringement is presented. This evolving legal landscape will likely influence future litigation involving AI and copyright law, with significant implications for both technology developers and content creators.
References
[1] https://www.bairdholm.com/blog/openai-gets-copyright-suit-dismissed-first-of-many/
[2] https://www.jdsupra.com/legalnews/no-harm-no-standing-for-now-dmca-8399331/
[3] https://www.linkedin.com/pulse/key-ai-developments-november-2024-holisticai-jpkhe/
[4] https://www.mckoolsmith.com/newsroom-ailitigation-2
[5] https://www.lexology.com/library/detail.aspx?g=cd34af38-1094-411c-9347-577004330a6e
[6] https://www.bsk.com/news-events-videos/no-harm-no-standing-for-now-recent-dmca-violation-claims-nbsp-against-openai-dismissed
[7] https://www.jdsupra.com/legalnews/recent-decisions-on-whether-ai-training-1219677/
