The debate over AI’s access to copyrighted content is heating up in the Delhi High Court, with a key legal expert warning that restricting language models like ChatGPT might do more harm than good. The concern? That limiting AI training data could inadvertently increase misinformation rather than curbing it.
In a case brought by ANI against OpenAI, Amicus Curiae Dr. Arul George Scaria argued that denying AI systems access to copyrighted materials could leave them reliant on a narrow pool of data, potentially distorting their responses. He told the court that allowing more access, rather than restricting it, might be the key to mitigating misinformation.
“The strength of large language models depends on the breadth of materials available for training,” Scaria noted. “Is it even feasible to build effective AI without tapping into copyrighted content? And if every small firm had to license content, could they even compete?”
The case raises fundamental questions about copyright law’s role in governing AI, with the court acknowledging the legal gray area surrounding generative AI. Four core issues are now under scrutiny:
- Whether OpenAI storing ANI’s copyrighted news data for training ChatGPT constitutes infringement.
- Whether ChatGPT using ANI’s data to generate responses qualifies as a copyright violation.
- Whether OpenAI’s use of copyrighted material falls under “fair use” exceptions.
- Whether Indian courts have jurisdiction, given OpenAI’s servers are based in the U.S.
Scaria contended that if AI training is for learning purposes—whether by a human or a machine—the law should not make a distinction. He also questioned whether AI’s “opt-out” mechanism for copyright holders is sufficient protection.
Meanwhile, ANI’s legal team countered that the company licenses its content to AI firms, arguing that OpenAI’s approach exploits its material without permission. The court acknowledged the dilemma: If AI companies must pay for every piece of content they use, does their business model even survive?
A second amicus, Adarsh Ramanujan, took a stricter stance, asserting that ChatGPT’s storage of copyrighted material alone is enough to constitute infringement unless it can prove fair use.
As the court wrestles with these questions, one thing is clear: the outcome of this case could set a precedent for AI development and digital content rights in India. The next hearing is set for March 10, and the battle over AI and copyright is far from over.