Chat Generative Pre-Trained Transformer (ChatGPT)  is an artificial intelligence (AI) chatbot that uses natural language to generate text in a conversational style. This language model tool created by OpenAI (a company specialising in artificial intelligence research) was launched in November 2022.

The emergence of ChatGPT has revolutionized the legal practice, allowing legal practitioners to perform their duties more efficiently. ChatGPT is able to generate legal articles and briefs that are the same as those written by  legal practitioners. However the use of this language model in legal practice raises concerns about potential copyright infringements, the dissemination of personal information and the accuracy of the information it generates.

When a legal practitioner uses ChatGPT to generate legal articles, they may potentially infringe on someone else's copyright, because ChatGPT is trained on a large corpus of legal documents, and there is a high risk of it reproducing copyright content without the user’s knowledge.

Another major risk posed by the emergence of ChatGPT in legal practice is the infringement of data privacy. Section 11 (1) (a)  of the Protection Of Personal Information Act 4 of 2013 (“POPIA”) prohibits the processing of personal information without inter alia consent from the data subject to whom the information belongs. In terms of the POPIA “processing” entails  the dissemination of personal information through transmission, distribution or making it available in any form. While it is standard practice for companies to inform their clients about the processing of their personal information, adding that information into ChatGPT may constitute further processing.

The risks of using ChatGPT in legal practice are not only limited to copyright and privacy infringements, it also raises concerns regarding the overall accuracy of the information it generates. This was evident in the case Parker v Forsyth NO and Others [Case No.1585/20][29 June 2023], where the plaintiff’s attorneys used ChatGPT to source legal cases, assuming the accuracy of its results. During the proceedings it was established that the cases referred to by the plaintiff’s counsel were entirely fictitious. The plaintiff’s attorneys relied on non-existent case law which was generated by ChatGPT without verifying its accuracy. The attorney’s failure to exercise due diligence led to the court ordering a punitive costs against the plaintiff. This case serves as an example  of the limitations of AI generated content and the need for human verification.

Whilst the use of ChatGPT and other AI related language models has many advantages, the consequences of its reckless use are dire. It is ethically expected that legal practitioners, approach legal matters with legally independent and questioning minds. Merely aping information generated by ChatGPT without verifying its accuracy undermines the integrity of the legal profession. It is therefore, imperative for legal practitioners to infuse modern technology with traditional research methods to ensure the accuracy and reliability of their work.

South Africa currently does not have any legislation regulating use of AI.  Although there are existing pieces of legislation  that can be applied in cases involving the usage of AI, the question remains whether the existing legislation adequately address the complexities of AI?

POPIA does not sufficiently address situations where personal information of a data subject is disseminated and/or obtained through AI outside of its predefined code.  Further, the Copyright Act does not adequately cater for instances where AI extracts and reproduces copyrighted content outside of its predefined code.

The question of how and to what extent should the use of ChatGPT and other AI related tools should be regulated is a challenging one. This thus makes it difficult to manage the relationship between the fast-paced emergence of AI tools and current regulations.  

Written by Nombuso Phiri – Candidate Attorney & reviewed by Matodzi Ramashia – Partner; Rams Attorneys