Jokes apart, the parent company of ChatGPT is now embroiled in yet another copyright infringement lawsuit, only this time, it comes from outside the Western hemisphere.

Indian news agency Asian News International (ANI), headquartered in New Delhi, is suing the tech giant for using its copyrighted content to train its large language models (LLMs), that power ChatGPT, without permission, reported Hindustan Times.

This lawsuit is the first of its kind in the country.

The AI company is facing accusations under sections 14 and 16 of the Copyright Act, 1957, along with claims of common law torts like unfair competition, unjust enrichment, garnishment and dilution. 

Sidhant Kumar Marwah, the lawyer representing ANI in the lawsuit, in a conversation with The Secretariat said there are three key issues in the case. First, that OpenAI is using ANI’s content to train its generative AI models, second, that ChatGPT is producing verbatim reports of ANI, which he says is "plain and simple copyright infringement". The third issue is that ChatGPT has also wrongly attributed things to ANI. 

“It (ChatGPT) claimed that Rahul Gandhi gave an interview to ANI. But in fact, Rahul Gandhi never gave an interview,” said Marwah. 

When large language models generate fabricated or inaccurate information that may be factually incorrect or nonsensical, the event is called 'hallucination'. Like most similar chatbots, ChatGPT warns users that its responses may not always be accurate. For instance, at the bottom of ChatGPT’s landing page, it notes: “ChatGPT can make mistakes. Check important info.”

What Does Copyright Law Say?

Under India's Copyright Law, except for very small exceptions like educational purposes or fair use, no one is allowed to use another party's copyrighted material for commercial gain, without obtaining a proper licence. Even publicly available content cannot be used without permission for the purpose of making money off it.

“Say you go to a cinema hall and watch a movie, right? It's in the public domain. It's a public place. What's stopping you from taking out your phone and recording a movie and then publishing it on your own? Similarly, why should ChatGPT be permitted, when an individual is not?” quipped Marwah.

Commercial use of copyrighted material requires a licence. Like any news wire agency, ANI has clients who typically pay for access to its wire services.

OpenAI has tied up with several news agencies and media outlets like the Associated Press, Axel Springer, the American Journalism Project, NYU, Time magazine, etc., to use their content to train its models.

In turn, OpenAI makes this information available through apps like ChatGPT, which is available for public use, but also has a subscription plan that offers advanced features to individuals and companies on payment of a fee.

“If people get the information straightaway from ChatGPT, they will not pay for my (ANI’s) subscription. Therefore, you're getting subscriptions at my cost, while I'm the one who's paying the money and creating the content,” said Marwah.

In a blog titled ‘OpenAI and Journalism', the company argued that training their AI models “using publicly available internet materials is fair use”. They also give the option for publishers to opt-out and prevent their models from accessing news websites.

“They (OpenAI) say that they are not journalists, they are not news organisations. Whereas, in fact, they actually are. Because they have entered into contracts with Financial Times, Conde Nast, etc., to distribute their content. So indirectly, you are a distributor of news content, and you're doing the same thing with my (ANI’s) content for free,” added Marwah.

The Secretariat learned that ANI had been in discussions with OpenAI since July 2024 over the unauthorised use of their content. But according to Marwah, OpenAI continued to use ANI’s content to train its models. It wasn’t until October 2024 that OpenAI confirmed ANI was added to the opt-out list.

This opt-out process was also adopted by the New York Times in August 2023 to prevent OpenAI’s tools from accessing their sites. Four months later in December 2023, the NYT filed a copyright lawsuit against OpenAI and Microsoft, alleging that millions of its articles were used to train ChatGPT. 

OpenAI currently has 13 lawsuits against it in the United States, two in Canada and one in Germany, since the inception of ChatGPT in 2022, but no injunction has been granted against the chatbot and no court has, as of yet, found copyright infringement in gathering of information by the model.

This was also the argument put forth by Amit Sibal, the lawyer representing OpenAI in the Indian lawsuit. Sibal argued that OpenAI doesn’t have any servers located in India, therefore there is no reason for the case to proceed in the country.

To this, Marwah clarified that under the Copyright Act in India, a copyright owner has the right to file a lawsuit in the jurisdiction where they reside, rather than where the defendant is located. This is an exception compared to other legal cases, where jurisdiction is typically based on the defendant's location.

“Facebook, Whatsapp, Twitter. None of these organisations have servers in India. All of them still have to comply with Indian law because they are providing services here,” added Marwah.

The Secretariat also spoke with Aman Taneja, an IP and tech lawyer at Ikigai, who explained that copyright law in India has primarily developed through judicial interpretations, making cases like ANI vs OpenAI a valuable opportunity for courts to clarify and refine existing provisions.

“Some countries have introduced what are known as text and data mining (TDM) exceptions to their laws to specifically enable AI training without obtaining explicit permission from copyright holders,” added Taneja.

Countries like the EU and the UK allow TDM for both commercial and non-commercial use, as long as proper credit is given. Japan has wide exceptions for machine learning and AI, and Singapore allows TDM unless there are contract restrictions. The US doesn't have a specific TDM rule, but its fair use laws often apply to these activities.

What Next?

The lawsuit will certainly influence the collision between intellectual property and generative AI, but is still a narrow part of an AI policy in India. “The provisions of the Copyright Act have not been interpreted by an Indian court in this context previously, so there are certainly interesting questions that the court will have to address in this case,” said Taneja.

The next step in the case is for OpenAI to submit its response, and for all parties involved to complete the necessary legal paperwork related to the interim injunction requested by ANI. The Delhi High Court on Tuesday refused to grant an interim injunction.

The court has also said that it will appoint an amicus curiae (Latin for "friend of the court"), who will present a neutral point of view and will advise the court on the law, and any technicalities involved. 

For now, once OpenAI submits its response and the formal documents outlining each party's arguments and claims are complete, the matter will be scheduled for review. The case will be heard next on January 28, 2025.