ChatGPT (Chat Generative Pre-trained Transformer) is a chatbot developed by OpenAI, an American AI (“Artificial Intelligence”) research laboratory. It has been launched recently, in November 2022. The fancy new bot has sent ripples across industries due to the vast array of functions that it can perform. The functions include writing code, composing emails, answering questions, engaging in conversation and even writing bedtime stories. There is much frenzy in digital and social media about potential job losses in the market arising from engagement of ChatGPT. However, this isn’t the only concern. Users and beneficiaries of the new AI may run into legal trouble.
In order to understand potential legal concerns that revolve around ChatGPT, one must first understand how it actually works. The new AI works by combining large amounts of data with fast processing and algorithms. Another key feature that it possesses is “automatic learning”, from data patterns. Unlike a conventional search engine, ChatGPT doesn’t merely depict the results available by browsing the web. It in fact uses the gallons of data in its possession to create and present content in response to a wide variety of requests.
The term “intermediary” is defined under Section 2(1)(w) of the IT (“Information Technology”) Act. It’s an entity that receives, stores or transmits electronic records on behalf of another person. Common examples of intermediaries are search engines, online payment sites, e-commerce entities and the like. The extent of OpenAI’s liability for content on ChatGPT depends largely on whether or not it can be classified as an intermediary. Clause 3(a) of Open AI’s terms of use vests responsibility for content on users. It in fact says that the input provided to ChatGPT belongs to the user. The terms thus indicate that OpenAI intends to abstain from claiming any ownership or liability for content generated through it. It may thus be inferred from OpenAI’s terms that it seeks to be classified as an intermediary
It is pertinent to note that an intermediary is only exempted from liability for data made available by it in certain circumstances, that are laid down in Section 79 of the IT Act. The intermediary can avoid liability if its role is limited to providing access to a communication system over which information is made available by third parties. An intermediary may also escape liability if it does not itself initiate a transmission, select the receiver of a transmission, or select or modify information contained in a transmission. ChatGPT modifies information fed to it by users to provide an output. It can even be said that it actually generates content upon receiving instructions from users. In the event of a dispute, courts may thus ponder upon the question of OpenAI’s intermediary status and liability.
In India, Section 13(1) of the Copyright Act, 1957 grants protection to copyright in original literary, dramatic, musical and artistic works, cinematograph works and sound recordings. Since ChatGPT creates content by using data that has been fed to it, many wonder whether it is infringing upon the copyright of other persons. Section 51 of the Copyright Act, 1957 stipulates what constitutes copyright infringement. Sub-Clause (i) of Clause (a) of the said section stipulates that a person can be held liable for doing anything, without a license, which can only be done by a copyright owner. Sub-Clause (ii) of Clause (a) of the said section holds a person liable for copyright infringement where they provide a place for communication of infringing work to the public. Prima facie, OpenAI may be held liable under both the sub-clauses.
Alternatively, one may argue that ChatGPT merely produces “derivative work”. Derivative works, also known as secondary works, are works derived from prior existing works. Open AI’s present terms of use stipulate that the input, or request made, to ChatGPT, is owned by the user. Further, the terms say that OpenAI assigns its right, title and interest in the output, or text generated in response to the input, to the user. The terms clarify that output generated may not be unique for each user and that the same output may be given to separate users. There is no question that the output generated by ChatGPT is derived from pre-existing data.
Copyright in derivative work has been discussed in extensive detail by the Hon’ble Supreme Court of India in Eastern Book Company & Ors. Vs. D.B. Modak & Anr. (2008)1SCC1. Law on the issue is fairly well settled by the said judgment. The new, derived work, must be somewhat different in character from the original work. Some degree of creativity must be employed. Even if the creativity is not strikingly apparent, it should be sufficient to give a new flavour to the derived work. In order to claim copyright in the derivative work, the author must exercise “skill and judgment”, not merely “labour and capital”. The Hon’ble High Court of Delhi applied the said judgment, in the case of Reckeweg v. Adven (2008) 38 PTC 308 and held that an author would not be entitled to copyright in a derivative work merely because efforts have been made to create it. Application of intellect would be essential for copyright to vest, as per the High Court.
The crux of the entire issue is whether “intellect” can possibly be applied by ChatGPT. The Oxford dictionary defines “intellect” as the faculty of reasoning and objective understanding, especially with regard to abstract matters. The word “abstract” is of great relevance here. ChatGPT has a fantastic set of abilities. It can create great content. Nevertheless, it cannot think in the abstract. Hence, one may find it difficult to concur with the proposition that it possesses intellect.
Foreign jurisdictions also restrict copyrighting of work that is mostly machine generated, with little to no human control on the form or expression of the work. The law on the said issue has been settled by the CJEU in Cofemel — Sociedade de Vestuário SA v G-Star Raw CV, C 683/17, wherein it has held that a work is only “original” if it adequately expresses the personality and creative choices of the author. The said decision was rendered by interpreting of a gamut of directives pertaining to copyright. In light of the same, the originality of ChatGPT’s output would be highly questionable. This is because the said output is generated by the capabilities of machine learning, with minimal human intellect involved. It is interesting that originality, as per the said decision, requires the “author’s creative choices” to influence the work and not merely any form of minimal human intervention.
We may also consider the English Copyright. Design and Patent Act. 1988, which vests authorship of computer generated work in the person making “the arrangements necessary for the creation of the work are undertaken.” Hence, a plain reading of the statute would imply that the author would be OpenAI, in case of ChatGPT. However, as per Clause 2(a) and (b) of Open AI’s terms of use for ChatGPT, its output is largely driven by the ‘input’ or the questions asked by the user. This further complicates authorship and copyright ownership for the content generated by ChatGPT in the UK.
The hindrances to authorship and copyright ownership are even more substantial under US Copyright laws. Presuming the nature of work produced by ChatGPT is derivative, in the US, creation of derivative work cannot take place without the consent of the owner of the original work. As per 17. U.S.C. §106(2), the creation of any derivative work is the exclusive right of the author. This explains the recent decision of the US Copyright Office, wherein it terminated the registration of copyright in a graphic novel produced substantially with the help of an AI tool, Midjourney
One school of thought, gradually gaining traction, subscribes to the view that AI, in the very near future, could evolve to such an extent that it may even have a mind of its own, or possess intellect. Nevertheless, the intellectual property regime in India and worldwide does not yet lay down IP rights for AI. Necessary amendments should be incorporated in this regard, to statutes that pertain to intellectual property.
Once such amendments are incorporated, the next issue to be tackled with respect to ChatGPT would be ownership of copyright in the content generated by it. We would recommend that OpenAI incorporates a feature which verifies that the content generated by ChatGPT is new and copyrightable, even if it is a derivative work. OpenAI should be held liable in the event that ChatGPT wrongly verifies blatantly copied material as a valid, derivative, copyrightable work. Subsequently, guidelines can be issued and regulatory frameworks can be brought about to further mitigate legal risks without stifling innovation.
It is strongly recommended that legislators, regulators and policy makers follow the above-mentioned recommendations. A balance can then be struck between phenomenal innovation and protection of the IP Rights of creators and artists.
www.grandmasters.in | 8 Years & Counting
www.rcls.in | 8 Years & Counting
www.itlegalsummit.com | 8 Years & Counting
www.bfls.in | 8 Years & Counting
www.maels.in | 8 Years & Counting
www.plcs.co.in | 8 Years & Counting
Copyright © 2020 Lex Witness - India's 1st Magazine on Legal & Corporate Affairs Rights of Admission Reserved
