New Frontiers of AI and Copyright Law: A Case Study of ANI VS Open AI

 

Whenever these two words—copyright law and AI—are heard together, the very first corollary drawn is with respect to the challenge of harmonizing the former with the latter. Copyright law is, by effect, applicable to works of an author, but nowhere is AI mentioned as an author, so to give AI the cloak of an author and AI's work the protection of Copyright Law was the very first question for consideration.

With the advancement of AI, such challenges have been identified that call for a significant overhaul in the current laws of Copyright and laws governing AI. While this challenge of harmony between work generated by Artificial Intelligence and Copyright Law is being attended to, a new challenge has come up calling for attention.

 

The Legal Battle

In the backdrop of advancement, AI has used copyrighted content to train itself without the author's permission. ANI, Asian News International, a prominent news agency that has a big role in providing multi-media news to various agencies, covers all news from all over India, discovered that Chat GPT, a prominent Large Language Model, an artificial intelligence chatbot that has the capabilities of generating human-like responses, was generating human-like responses from sources of ANI's news articles without authorization. This led to ANI filing a copyright infringement suit before the Delhi High Court against Open AI, a leading AI research company.

ANI, therefore, filed a copyright infringement suit before the Delhi High Court against Open AI. The case, which is complex and multifaceted, revolves around the unauthorized use of ANI's content by Open AI's Chat GPT. ANI firmly held its ground, asserting that Open AI should have obtained permission or respected its licensing agreement. In contrast, Open AI, as the respondents, challenged the jurisdiction of Indian Courts and argued that their training for Chat GPT is done from publicly available data, thus no copyrighted content is used.

 As admitted by the counsel on behalf of Open AI, Open AI is currently facing suits of similar kinds in sixteen different countries, which shows the gravity of the problem.

A parallel case worth mentioning is The New York Times lawsuit against Open AI in 2023 concerning data scraping and unauthorized content use. While that case proceeds in the U.S., the ANI case presents an opportunity to observe how the Indian legal system will respond to these emerging concerns. This response could potentially shape the future of Copyright Law with respect to Artificial Intelligence, leading to new regulations and practices.

 

Investigation Bottlenecks

What are LLMs? Large Language Models are a type of foundational AI model developed using a huge dataset, enabling them to comprehend and produce human-like language and other content formats across diverse tasks. What number is a considerable amount? LLMs can study approximately 3.10 billion pages of data at a single time. This number is not as huge as it can get; it's bound to increase with the advancement. With such a large number of data to study from, it would be next to impossible for anyone to know what a model like Chat GPT is learning on the internet to find copyrighted content being studied.

Furthermore, after studying all the data and learning like a human, LLMs tend to rephrase what they have studied while answering a particular prompt. This rephrasing, akin to human explanation, makes it a complex and resource-intensive task to backtrack and analyze the response, especially when it involves copyrighted content. This complexity adds another layer of challenge to the legal battle between AI and Copyright Law.

____________________________________________________________________

~ Purav Garg

Research Associate, World Cyber Security Forum

(B.Com LLB (H), UILS, Chandigarh)