The blog has been authored by Vir Malik, a 3rd-year student at Gujarat National Law University (GNLU), Gandhinagar

With the rapid influx of technology, the world’s legal landscape has had to constantly evolve to keep up with its pace. One such prevalent issue is Artificial Intelligence and the unexplored and under-legislated threats that it may pose to many sectors. Recently, the Calcutta High Court heard and ruled on a matter regarding ChatGPT and its alleged discriminatory practices against a privately owned company, IndiaMart. This paper will address how the judgment of the case at hand produced a new classification for AI models, and how that classification opens up a can of worms that seem to be ahead of India’s proposed legislative schedule, compelling us to act fast.

The Case

IndiaMart is a business-to-business marketplace that has been around for the last 30 years. The case started with IndiaMart seeking interim relief against OpenAI. This was brought about because of the allegation that ChatGPT had systematically excluded IndiaMart’s products and services from generative responses given to certain prompts. This issue was intensified with the knowledge that ChatGPT allegedly showcased biasness by providing users with links to directly competitive platforms. IndiaMart’s largest argument rested on the fact that ChatGPT has become substitutive of search engines in today’s age. This overt reliance should then lead to its classification as an “Intermediary” as is defined under section 2(1)(w) of the IT Act, essentially meaning a “person” that transmits information or provides a service based off of a record received from another “person”, which in turn would make it subject to Rule 3(1)(n) of the IT rules act. This rule and classification prohibits discriminatory behavior from intermediaries, which leads to the invocation of Articles 14, 19 and 21 of the Indian Constitution. This framed ChatGPT’s act of exclusion as a clear violation of rights to equality and trade. It is evident that the pure power AI holds in our world across every sphere is expanding at a very high rate. The fact that a company can put such heavy allegations on an online platform due to a real negative impact on their brand speaks volumes about technological advancements and the interchangeableness that these language models have been able to access.  

In response to the arguments laid out, Justice Ravi Krishan Kapur laid out a three-pronged approach with his judgment culminating in a judgment heard around the nation. Contrary to the “Intermediary” classification by IndiaMart, ChatGPT was touted to be an “originator” under section 2(1)(za), something that is too premature of a classification given the little legislation Indian law has on this. The simple reasoning was that ChatGPT could not be called a search engine as search engines are only used to rank and present pre-existing content in its rawest form, not a collation of carefully picked and algorithmically synthesized content tailor-made for the user. The court recognised that the IT Act came into enactment in a pre-AI world, where the origination of responses was something not conceivable by technology at the large and accessible scale as seen today. The court also acknowledged that applying those old definitions to a language model is a “Complicated and vexed question of both law and fact”, requiring expertise and perhaps a reworking of many existing statutes. However, one thing was said to be of immediate effect: if ChatGPT is beyond reasonable doubt not classified as an intermediary, it is not bound by rule 3(1)(n) or, for that matter, the entirety of rule 3 of the IT rules.

The Immediate Impact

This classification doesn’t simply impact a private company’s appeal for advertisement; it changes the legal rights accessible to all large language models in today’s world and those to come. Since AI models now, temporarily at the very least, have a precedent to be called originators, most AI outputs will be regarded as original content. This fails to look at how this “original” content is often the result of a series of carefully crafted prompts that compel the language model to give the user their desired output.

This ruling will also deprive AI platforms of the safe harbor provided under Section 79 of the IT Act. This would completely disrupt how AI companies launch their LLMs, since newer models are becoming increasingly autonomous in nature, which would border on the assumption of being categorized as an “originator”. Newer models, such as Claude’s Fable 5 (if it gets clearance to be released outside the USA), will either have to be prepared for the chance of incoming litigation or heavily strip their model of its powers. This, in turn, would lead to a plethora of charges being deployed towards LLMs and AI companies, as they are against human originators. Charges like sedition, defamation, discrimination and even abetment must be accounted for, reflecting the liabilities of an originator, and lifting the corporate veil would become easier than ever, opening a floodgate of issues.

A Path To Resolution

There are a variety of ways to tackle this problem. To start with, a clear regulatory path would be the creation of a distinct, third category of “AI content provider”. Firstly, this removes the hassle of an originator being a “person” as per section 2(1)(za) with a separate category of AI content providers labelling and rendering themselves unique, which significantly reduces the various liability facets one might have to face. In essence, this means that liabilities towards companies like OpenAI will be determined by a definition and not inferred.

Another way is to amend section 11(c) of the IT Act. This section talks about how an electronic record is to be attributed to the originator by an information system programmed by or on behalf of the originator to operate automatically. If it were tweaked to clarify that human initiation of a prompt does not make the AI an “originator” since there is a clear pathway to arrive to a conclusion, it would make for a much more equitable distribution of liability should any situation arise due to the recategorisation.

Possibly, the most important way to combat this issue would be to try to build a risk-based liability Schedule for AI-generated content. Now, treating every single generative system as a full-service originator, or an intermediary would impose unbound and severe liability that would be disproportionate to the risk that should be dealt with on a case-by-case basis. The EU Artificial Intelligence act’s four-tiered structure provides a separation of systems into: “Unacceptable risk”, “High risk”, “Limited risk” and “Minimal risk” systems. These classifications stem from the intended purpose of usage of the systems, not the technology running them. The same technology could be used to make school projects while also framing contracts with expert precision. As we see with other structures in Indian legislation, “unacceptable risk” forums and their equivalents, like “top layer” NBFCs, are oftentimes at the discretion of the Government, with likely no organization falling under that category. This showcases that India has previously and continues to adopt a categorisation  model that would be extremely useful for the scenario at hand. If India could adopt this model, to understand liability distribution and the line between an intermediary and an originator, and to what varying degree an originator should bear liabilities, it would be an ideal and a big step towards building around the Calcutta ruling.

Conclusion

In conclusion, the “originator” classification by Justice Ravi Krishan should stir up a domino effect, and hopefully a wake-up call for Artificial Intelligence legislation. The classification of OpenAI to an “originator” resolves nothing, and while exposing a glaring hole in legislation, it also opens up a sinkhole that could lead to less interest in Indian expansion for certain AI models. However, with proper strategizing and a smart approach towards categorising and layering, India could see a massive improvement in their system of governance. The need of the hour, the talk of the times and the focal point of every headline are two words: Artificial Intelligence. The quicker India can keep up with the ever-evolving state of technology, the better it is for us.