[This is the first part of a two-part article analysing ChatGPT and its legal implications. It is authored by K Nand Mohan in the second year, and RS Sanjanaa in the third year at Symbiosis Law School, Pune. The second part can be found here.]
Amidst news of computer scientists from University College London having developed an artificial intelligence ‘judge’ and the world’s first “robot lawyer” defending an alleged traffic rule violation, questions regarding the impending ‘robotification’ of the law and its constituents have once again come to the fore. At the vanguard of this has been OpenAI’s ChatGPT, which has made the headlines ever since its release on the 30th of November, 2022.
ChatGPT is an Artificial Intelligence (“AI“) system that uses huge text data, from pre-existing sources including books, articles, and websites, to learn natural language sequences and patterns to produce complex responses from simple input. To this end, it employs a modified version of the neural-network machine learning model (“ML“) called “Generative Pre-Trained Transformer” (GPT). ChatGPT, like any other revolutionary technological advancement, presents a host of new legal questions that might have far-reaching consequences on the various components of the legal system. This article appraises the conceivable applications of the AI system in the legal domain, along with its legal implications in detail, while determining the need for an AI-specific regulation in India.
Legal Use Cases
A. Legal Research and Advice
ChatGPT may be employed for efficiently analysing pertinent case law, laws, and other forms of legal literature. Predictive analysis of litigation outcomes might be provided using ChatGPT, allowing lawyers to render better advice regarding the outcome of a matter, and to decide whether to take it to trial or settle it out of court. Further, it also helps clients determine whether the lawsuit is financially viable.
In reference to the public, justice inequality or disproportionate access to legal representation and information is another aspect that ChatGPT may address. A reputed law firm, Linklaters, tested the dependability of its legal advice by feeding it 50 prompts and concluded that legal advice is typically context specific and depends on various external circumstances.
However, while the AI system produces bold and authoritative answers, it fails to pick up on nuances. Some users have reported made-up statutes or incorrect legal information, placing doubts on its reliability. With more data and smarter learning algorithms, future AI/ML solutions might come very close to picking situational aspects as humans do, at least in formalized empirical areas. Besides, when paired up with tools that specifically cater to legal research, such as Ross Intelligence’s API, the competency of the AI system can potentially overcome its inherent limitation.
B. Legal Drafting
The ability of ChatGPT to draft contracts and whip up legal documents is being widely spoken about. When asked to draft a Non-Disclosure Agreement under the jurisdiction of New York of 1000 words, the chatbot presents a ‘template-like’ result which could help attorneys construct specific contract provisions if prompted properly. Its application is flexible to numerous domains of law and addresses questions of family law, employment law, etc., in a generic yet prompt-specific manner. The OpenAI software can also review a contract and point out any mistakes while also suggesting changes to bring it into compliance with legal language and terminology.
Ryan Sami of Standard Draft suggests the use of standard contracts with a more methodical approach to developing and executing them, instead of AI-generated contracts, which are limited to a small dataset. In theory, the software might be used to generate preliminary versions of items that don’t call for much original thought. Nonetheless, it is likely to be misleading in more complicated and nuanced legal documents since it fails to correlate facts to the law, understand the law, or use human skills like persuasion and emotional intelligence. ChatGPT could, however, benefit from tying up with other softwares such as Kira System’s Machine Learning Software, ‘Kira’, that aids in accurate and efficient contract and document content identification, extraction, and analysis.
C. Legal Marketing
Specific to lawyers and firms is ChatGPT’s potential in the legal marketing field. ChatGPT can be integrated into the firm’s website or utilised on chat platforms to interact with prospective customers in real-time and reply to their queries or reviews. Further, it can schedule appointments through calendar integration and can be utilised for content writing and creation. When asked about the four elements of negligence for a personal injury claim, the chatbot gave a response comparable to those found on a law firm’s website, despite being quite generic and superficial.
D. Legal Academia
Legal Academia is another prominent area where we will see an impact of ChatGPT. The chatbot produces original, paraphrased text in a matter of minutes – whether your assignment is to write a theatre script or to advise a relative on internet security and privacy. It circumvents plagiarism checkers such as Turnitin and recently, it even took the bar exam where it outperformed random guessing.
It is worth noting, however, that a Princeton student has developed an app named GPTZero that can determine whether an essay is produced by a human being or AI in a short amount of time to address plagiarism concerns by professors.
[The second part of this post analyzes the legal implications of ChatGPT, and suggests a framework for its regulation]