Introduction
The Delhi High Court’s interim ruling in the case of ANI Media Private Limited v. OpenAI OpCo LLC may become a pivotal point in the legal landscape surrounding Generative AI in India. While the case has not yet resolved the legality of AI training, it has temporarily placed the training of commercial Large Language Models (LLMs) under the fair dealing exception of ‘personal and private use, including research’ as outlined in Section 52(1)(a) of the Copyright Act, 1957. The court, in denying an interim injunction against OpenAI, stated that training LLMs falls within this exception and therefore does not constitute copyright infringement as per Section 51 of the Act.
Interim Rulings and Their Influence
While the judgment is interim, such rulings often set a precedent that influences future legal interpretations, potentially broadening the application of Section 52(1)(a). This judgment appears to stretch the text’s limits, reversing legislative history, and sidestepping Indian precedents while implicitly adopting the American fair use doctrine, which the court admitted has limited relevance in the Indian context.
Redefining ‘Private Use’ and Its Implications
The court’s interpretation of ‘private’ in the context of ‘private and personal use’ extends to corporations, citing dictionary definitions that include ‘a person, group, or class.’ OpenAI’s use of ANI’s works in a ‘closed space without public access’ was deemed private by the court. This challenges traditional interpretations where ‘private’ and ‘personal’ are viewed as overlapping, requiring a coherent interpretation. The ruling’s logic suggests that massive, industrial-scale data usage by corporations could be justified under ‘private use,’ potentially undermining the original intent of protecting individual use.
The Evolution of Legal Definitions
Historically, the term ‘private study’ was amended to ‘private use’ in 1994 to protect individuals from harassment over casual copying. In 2012, ‘personal’ was added to reinforce individual protections. The legislative intent was to bolster individual rights rather than create a corporate exemption for large-scale copying. Copyright law prohibits unauthorized copying, even if never shared publicly. However, according to the court, corporations can retain copies privately without infringing, effectively nullifying the rule against copying.
AI Research and Copyright
The court deemed LLM training as research, applying the ‘doctrine of updating construction’ to adapt old definitions to contemporary contexts. However, this doctrine is typically reserved for evolving methodologies, not fundamentally different activities. Traditionally, ‘research’ involves a person studying a text to gain knowledge. The court’s interpretation contradicts this by suggesting that AI’s internal data usage for training can qualify as research.
More Read
Commercial Use and Legal Precedents
The court noted the absence of the term ‘non-commercial’ in Section 52(1)(a), implying commercial use is permissible. This interpretation diverges from previous rulings, like Jiwan Publishing House and TIPS v. Wynk, where commercial intermediaries were not allowed to claim end-users’ research as their defense. Although OpenAI claims internal use, it parallels the actions of companies previously found to infringe copyright by making unauthorized copies for profit.
Access Does Not Imply Authorization
The court held that since ANI’s content was freely accessible online, OpenAI’s copying was not a violation. This stance conflates accessibility with authorization, a perspective not supported by Section 51 of the Act. The court shifted the burden to ANI, suggesting they should have technologically blocked OpenAI’s access, introducing an opt-out requirement absent in Indian copyright law.
Financial Harm and Licensing
The court’s assertion that OpenAI’s actions caused no financial harm to ANI overlooks the existing market where licenses are purchased for such uses. The court’s decision to leverage ANI’s license negotiations against them could deter creators from seeking fair compensation.
Conclusion
The ruling redefines Section 52(1)(a), crafted to protect individual readers and researchers, into a broader corporate exemption. This change could devalue Indian news publishers’ data assets and disrupt the AI ecosystem it aims to protect. While the court acknowledged the public interest in AI development, it could have reached a similar interim decision without fundamentally altering copyright law principles. The question remains whether this constitutes interpretation or legislative action.
