ChatGPT, AI content: All about legal challenges pertaining to copyright under Indian law

The days of Googling for specific information and going through several results are numbered with the recent developments in artificial intelligence technology, where platforms like ChatGPT dish out content regarding specific questions.

ChatGPT, developed by OpenAI, is an AI model designed for conversational interactions. It can answer questions, offer explanations, generate text, and engage in interactive text-based conversations across a wide array of subjects.

The introduction of ChatGPT has reignited discussions about the intersection of AI-generated content and copyright, highlighting several challenges. Existing copyright laws do not seem well equipped to provide copyright registration or protection for AI-generated works in India, nor do they explicitly acknowledge AI as an author.

Even ChatGPT itself acknowledges the complexity of content ownership under Indian law when it comes to AI-generated materials. The legal framework in India lacks specific provisions addressing the ownership of AI-generated content.

Two years ago, the Parliamentary Standing Committee recommended the creation of a distinct category of rights for Artificial Intelligence and related innovations, addressing their protection as intellectual property rights.

In June of this year, the Minister of State for Electronics and Information Technology, Rajeev Chandrasekhar, expressed the government’s intent to regulate artificial intelligence to safeguard “digital citizens.”

It is evident that numerous issues arise when attempting to fit AI-generated content within the copyright framework, including questions regarding ownership of such intellectual property.

To delve deeper into these matters, India Today interviewed Pravin Anand, Managing Partner at one of India’s top IPR law firms, Anand and Anand. Regarding whether ChatGPT can infringe on original content, Anand noted that it is possible that content generated by Generative AI tools, such as Large Language Models (LLMs), may infringe on the copyrights of third parties, contingent on the specific circumstances of each case.

Is it possible to get copyright for AI-generated content?

As for whether AI-generated content can be copyrighted, existing copyright laws stipulate that the first owner of copyright in a work is the author. India’s Copyright Act of 1957 does not specifically address AI-generated works or recognize AI as an author. One key constraint in copyright protection for AI works is that they must be original and creative to qualify for copyright protection.

Originality is a benchmark used to determine copyright protection for a work. Section 13 of the Indian Copyright Act states that copyright exists in “original literary, dramatic, musical, and artistic works.” However, the Act does not explicitly define “originality,” leaving it to the courts to determine whether a work meets this criterion.

Content generated by AI may not meet the standard of originality or creativity because it relies on data from existing sources on the internet and data provided during training.

‘Computer generated work’: The Copyright Act in India was amended in 1994 to include computer-generated works, including literary, dramatic, musical, or artistic works. Section 2(d)(v), was introduced in the act to define the authority of such works as “the person who causes the work to be created.”

How the term “person” is defined and interpreted here becomes relevant, since as of now only natural persons have been recognized as authors under law. What therefore also needs to be clarified by the law and by courts is the legal status of AI – whether AI can be defined as a ‘person’ under the law, and if yes, to what extent?

While dealing with a case of copyright claim by CBSE over question papers, Delhi HC had said that authorship can only be attributed to a natural person, and copyright couldn’t be asserted without evidence of individual involvement in the framing of those question papers.

In another case, Delhi HC had refused to recognize copyright claim over computer-generated list, due to lack of human intervention.

In a rare incident which had hogged attention, an AI-based app ‘Raghav’ was recognized as co-author of a copyrighted work. However, later the Copyright office had objected to the same and sought to cancel the registration. While application to register AI (RAGHAV) as the sole author of the work was rejected, the Indian Copyright Office had allowed the application where the creator was named a co-author to the AI ​​tool.

Can AI and ChatGPT content be a threat to rights to privacy & publicity?

Anand also discussed how AI-generated content can infringe on an individual’s right to privacy and publicity. He cited a recent case in the Delhi High Court, Anil Kapoor v. Simply Life India & Ors., where the court ruled that using generative AI tools to depict famous personalities in fictional scenarios violated their personality rights.

According to Anand, while using Generative AI for innocent, personal or other bona fide use is not a problem, using it to portray an individual without her consent, and then using the output (such as a photograph or a video) for commercial purposes, is unlawful.

Who is the ultimate owner of content generated by ChatGPT and other AI platforms?

Anand also told us that this would depend on the policies and terms of each platform. In the case of ChatGPT, its terms and conditions state that while the content generated by ChatGPT belongs to the platform, it is assigned (ie, ownership is transferred) to the user, whose prompts led ChatGPT to generate the content.

He added that, the platform also clarifies that if ChatGPT gives the same output to two (or more) users, then neither has right over the content. For example, if the response to a question “What shape is an egg” is “The shape of an egg is oval”, and this response is given to two or more users, then neither can claim rights in the answer.

As per ChatGPT’s terms and conditions, this is “due to the nature of machine learning, Output may not be unique across users and the Services may generate the same or similar output for OpenAI or a third party”.

Creative Writer’s Copyright issues: India Today also spoke to Anand about the cases being filed by writers and creative artists against companies that are creating and operating AI powered LLMs (such as OpenAI).

Anand said that many such cases have been filed recently in the USA, but there is no decision on these aspects yet. Besides traditional copyright instances some of the more interesting and nuanced issues raised by authors in their lawsuits against LLMs are:

Source of training the LLMs – That these LLMs are being trained on datasets, which comprise the authors’ works, without taking any permission from the authors.

• Systematic theft: Many a time, the datasets may contain pirated versions of the authors’ books.

• Moral rights violation: The ability of LLMs to produce modified or alternate versions of the authors’ works, and thus, creating an output which may denigrate the author and his original work.

• Endangering livelihoods and unfair practices: LLMs can be increasingly used to generate content which would otherwise be given to authors, against an appropriate fee.

• Royalty payments: Authors argue that LLMs, which generate output based on their works, should pay a reasonable royalty to the authors.

Copyright law provisions that may protect AI content

The use of copyrighted content in by AI tools is also defended on the basis of the doctrine of fair use or fair dealing under the copyright law. Section 52 of the Indian Copyright Act, provides for certain exceptions to infringement of copyright, and permits limited use of copyright material without the owner’s authorization.

Although not specifically defined under the law, fair dealing with a literary, dramatic, musical or artistic work for the purpose of research or private study or criticism or review, may not constitute an infringement of copyright.

Advocate Pravin Anand told us that the test for copyright infringement in India is to see if the content is “substantially similar” to the copyright protected work, and if it is an exception to infringement under Section 52 of the Copyright Act, 1957.

According to him, the literary works (text generated by these LLMs), and its use for private consumption, research, review or educational purposes in schools or universities does not constitute infringement. On the other hand, if an output generated by LLMs based on prompts given by third parties, is used for commercial purposes, or for any other purpose that is not covered under the exceptions to infringement, then it may constitute copyright infringement.

For example, if a person prompts an LLM to produce a translation of a book, or make an alternate ending to the story in a novel, and publishes it in his or her own name, then this will infringe the copyright of the author of the book.

The classification and determination of what falls under fair dealing has obviously been something that is decided by Courts, keeping the facts and circumstances and certain guidelines in mind.

Another remotely explored aspect is also the possibility of the AI ​​content being protected for being derivative work”.

Derivative works are based on other works already protected by the intellectual property rights, and may be protected under the law of intellectual property, if there are variations from the original or parent work. The responses from ChatGPT, which is a form of Generative AI, are based on the model’s learning based on the data fed to it and multiple pre-existing sources.

Derivative works can be generated either with the explicit permission of the copyright holder or by using works that are in the public domain. To qualify for copyright protection, a derivative work must introduce a significant degree of alteration to the original material, and the degree of alteration required varies depending on the type of work involved.

For example, in the case of some works, simply translating them into another language may suffice, whereas others may need a complete shift to a different medium. In essence, a mere substitution of a few words in a written work, for instance, is insufficient to create a derivative work; a substantial modification of the content is imperative. Similarly, a derivative work must integrate enough of the original material for it to be evidently rooted in the original source.

Conclusion

The emergence of ChatGPT raises significant intellectual property concerns. Amendments to copyright laws may be necessary to address the unique challenges posed by AI technology. The legal implications of using such tools will likely remain complex and uncertain until more definitive laws are established.

With input from Srishti Ojha