ChatGPT: Will the AI legal framework come of age in India?

The AI-driven tool is built using OpenAI’s GPT-3 family of large language models and fine-tuned with both supervised and reinforcement learning techniques. (Illustration by Suneesh Kalarickal)

Recent discussions around the fourth industrial revolution on mechanisation and the pervasive use of artificial intelligence (AI) and machine learning to displace human resources have necessitated that legal regulation catches up with such usage of AI. While the government of India has released several reports and policy documents outlining the contours around the use, liability, and responsibility of utilising AI-driven technologies, it is imperative that an effective framework is put in place to bridge the regulatory lacunae in the use of ground-breaking technologies such as ChatGPT which has reached 100 million users just two months after its launch.

Chat Generative Pre-Trained Transformer or ChatGPT is a chatbot launched by OpenAI, the non-profit wing of the American artificial intelligence research laboratory, in November 2022. The AI-driven tool is built using OpenAI’s GPT-3 family of large language models and fine-tuned with both supervised and reinforcement learning techniques. Trained using data about code and existing information from the internet to engage in dialogue and human-like responses, ChatGPT can create content such as poems, stories, software code, compose music, and so on.

On being asked to describe itself, ChatGPT responds with – “ChatGPT is a large language model developed by OpenAI. It is trained on a massive dataset of internet text and is capable of understanding and responding to a wide range of natural language inputs. It can be used for a variety of tasks such as language translation, question answering, text completion, and more. It uses the transformer architecture and deep learning to generate human-like text”. Naturally, using this response ‘as is’ in an original article is tempting, but questions as to the authorship of this response remain unanswered.

AI As A Person

Under the Copyright Act, 1957, a person who causes literary, musical, artistic, or dramatic works, to be created using computer-generated programmes, is designated as an 'author'. The legal definition of the term, 'person' covers artificial and juristic persons. However, in India, AI has not been granted the status of an artificial person – much in contrast to other countries, for example, Saudi Arabia which granted citizenship to the AI-driven robot – Sophia, recognising it as a legal person. Countries such as the US that have dealt with the registrability of AI-generated works require 'human intervention' for such works to be granted copyright registration. In India, when a copyright registration request was made for an artwork created using the 'Raghav Artificial Intelligence Painting App', India granted co-authorship to the applicant, Ankit Sahni and the AI app. However, the copyright office later withdrew the co-author status, reasoning that AI does not constitute a legal person. Authorship by AI vests entirely on how a 'person' is construed by the Indian legislature and courts.

Similar to the US, India recognises the 'sweat of the brow' and 'modicum of creativity' doctrine to determine if a piece of work is copyrightable i.e., if a [Eastern%20Book%20Company%20v.%20D%20B%20Modak,%20(2008)%201%20SCC%201%20(India)]certain level of human skill and intervention was utilised along with a minimal degree of creativity. AI can create both original and derivative works using the information used to train AI and built-in algorithms that allow it to learn from data inputs. To that extent, the doctrine of 'modicum of creativity' applies to works created by AI-powered tools such as ChatGPT. However, the registration of such works will finally vest on the question of authorship by AI – an unresolved question right now in Indian jurisprudence.

Copyright Ownership

On the subject of liability of content produced by ChatGPT, since ChatGPT is generally used like a search engine – where input is fed by users into the AI tool, it may fall within the ambit of an 'intermediary' as defined under the Information Technology Act 2000. Therefore, it will be entitled to the safe harbour protection statutorily available to intermediaries.

On liability for the content used to train ChatGPT, there is of course ambiguity on how such content has been secured and the remuneration for such access. But as usage of ChatGPT increases, it’s likely that questions around copyrighted material used to train it will surface. For example, Stable Diffusion (a latent text-to-image diffusion model capable of generating photo-realistic images given any text input) was recently sued by three artists for being trained using copyrighted artworks of artists who did not permit such use.

As people use and interact more with ChatGPT, OpenAI is likely to collect more personal data, helping ChatGPT to offer user-specific responses. Currently, there is limited information available on the privacy models adopted by OpenAI, however, reportedly it has entered into data protection agreements to overcome concerns on data retention and extraction with corporates who have shown interest.

Commercial Implications

The commercial implications of ChatGPT on professions relying on content and information, such as legal practice and journalism could be beneficial to the extent that ChatGPT could be used to analyse and organise documents, assist in research, translations, predict outcomes of legal cases or social occurrences with similarities in history or jurisdictions. As an upside, this could create and enhance the value of human creative input and may create further opportunities for AI development and implementation. However, on the downside, tools like ChatGPT could adversely impact knowledge workers (to the extent that it could replace humans for tasks such as research, report writing, and drafting of documents), and the learning and cognitive abilities of students accessing a tool programmed to perform simple tasks. Interestingly, recognising this ability of ChatGPT, Bengaluru-based RV University recently banned university students from using ChatGPT for exams, assignments, and lab tests.

Undeniably, AI can perform certain tasks more efficiently than humans, and given this, it may be too early to predict how ChatGPT will specifically impact our workforce. However, considering its varied implications across different workstreams, it is likely to lead to an acceleration of not only different AI models, but also the evolution of AI and ChatGPT-based legal frameworks in India and across other jurisdictions. This especially includes laws relating to the ownership of intellectual property, content liability, the liability of platforms for their services or software products, and data privacy.

Tanu Banerjee is Partner and Garima Kedia is Associate at Khaitan & Co. Views are personal and do not represent the stand of this publication.