Examining copyright infringement and liability in Generative Artificial Intelligence training and use: a legal perspective in South Africa and beyond

Once again, humanity has welcomed technological advancement, this time around artificial intelligence, with mixed reactions. The creative industry is no exception to this rapidly evolving technology, with generative AI (genAI) deepening its claws in the creative industry. What lies within the fabric of genAI is a primary concern. In order to be trained (taught), genAI ingests enormous amounts of data, which is harvested indiscriminately. This is a cause for concern for those whose work is harvested and utilised without compensation, credit and consent. In addition, when genAI is deployed, the user's input prompts it to create works of their desires, ranging from images to musical lyrics. To that end, it remains to be answered whether the use of works for the purpose of training genAI and the generation of works by genAI trained using copyrighted works amount to copyright infringement. It is the duty of this dissertation to examine whether the South African copyright regime would deem it an infringement to make use of copyrighted works for training genAI. In addition, this dissertation goes further to examine whether there is a possibility of copyright infringement materialising when a user generates works through genAI. In this dissertation, the training of genAI will be referred to as the “input phase”, while the use of genAI by a user will be referred to as the “output phase.” In addition, the issue of who is liable when copyright infringement materialises will be analysed. In the end, the author submits recommendations for South Africa to address copyright infringement liability. Leading jurisdiction in copyright and AI regulations will be infused to enrich the discussion

Grantor dc:publisher.institution

School For Advanced Legal Studies

Year dc:date.issued

2025

Author dc:creator

• Mikioni, Tendai

Advisor dc:contributor.advisor

• Ncube, Caroline

• generative AI
• artificial intelligence
• South Africa

Language dc:language.iso

en

Handle dc:identifier.uri

http://hdl.handle.net/11427/42670

OAI identifier oai:identifier

oai:open.uct.ac.za:11427/42670