ANI vs. OpenAI: Amicus Curiae Cautions That Limiting AI Access To Copyrighted Content Will Aggravate Misinformation
The Delhi High Court heard arguments from Amicus Curiae Dr. Arul George Scaria in ANI’s copyright lawsuit against OpenAI;
ANI vs. OpenAI: Amicus Curiae Cautions That Limiting AI Access To Copyrighted Content Will Aggravate Misinformation
The Delhi High Court heard arguments from Amicus Curiae Dr. Arul George Scaria in ANI’s copyright lawsuit against OpenAI, where he asserted that restricting Large Language Models (LLMs) like ChatGPT from accessing publicly available copyrighted content could worsen misinformation. Scaria, a professor at the National Law School of India University (NLSIU), Bengaluru, emphasized that increasing access to diverse materials is essential to counter misinformation, as limiting training data could exacerbate the issue.
During proceedings before Justice Amit Bansal, Scaria questioned whether it was even feasible to develop LLMs without access to copyrighted materials. He highlighted that the effectiveness of such models depends on the breadth of their training data. He also pointed out that while OpenAI is a major tech company with substantial financial resources, smaller firms would face significant challenges in obtaining licenses from all copyright holders to enter the market.
Scaria argued that the concerns raised in the case extend beyond private copyright claims and should be viewed as matters of public law, suggesting that copyright law may not be the appropriate legal framework to address such issues. The Court acknowledged the absence of specific legal provisions governing generative AI models.
ANI has alleged that OpenAI is using its copyrighted news content without permission for commercial gain and to train ChatGPT. The Court, which had previously appointed Scaria and Advocate Adarsh Ramanujan as amici curiae, is now examining four key legal questions:
1. Whether OpenAI’s storage of ANI’s news data for training ChatGPT constitutes copyright infringement.
2. Whether generating responses using ANI’s copyrighted content infringes the news agency’s rights.
3. Whether OpenAI’s use of ANI’s data qualifies as "fair use" under Section 52 of the Copyright Act, 1957.
4. Whether Indian courts have jurisdiction over the case, given that OpenAI’s servers are located in the United States.
On the issue of storage, Scaria contended that the law does not differentiate between machine and human learning, and storage—whether temporary or permanent—should be permissible if used for learning purposes. He also emphasized the importance of considering the ultimate purpose of data usage when assessing liability. Addressing the question of infringement, he argued that mere similarity between LLM outputs and copyrighted material should not automatically lead to infringement claims without examining applicable fair use exceptions under Section 52.
Scaria further urged the Court to avoid rigid interpretations of copyright law and to assess whether OpenAI’s opt-out mechanism provides adequate control to copyright holders. He also raised a broader question: whether major news outlets like The Hindu or Indian Express would stop subscribing to ANI simply because ChatGPT provides similar information.
ANI’s counsel, Advocate Sidhant Kumar, countered that the agency already licenses its content to LLMs. In response, the Court questioned the sustainability of OpenAI’s business model if it were required to pay licensing fees to all content providers.
Amicus Curiae Adarsh Ramanujan then began his submissions, stating that his perspective differed from Scaria’s. He argued that ChatGPT’s storage of copyrighted material itself amounts to infringement, asserting that copying is an act of infringement unless explicitly covered under fair use.
The matter has been scheduled for further hearings on March 10,