Artificial intelligence (AI) is becoming a powerful tool for research, and automating information processing tasks that previously required human attention. Creating an AI model involves input data, processing and output. Both input data and output raise new questions for copyright and the creative processes that copyright is intended to support. Does use of a copyrighted text or an image in training an AI model infringe copyright in that text or image? Legal uncertainty on the use of images, text, and the like to train AI will hamper development of AI in Africa. AI trained in the Global North already excludes Africa both for input data and the ensuing outputs so it is essential for Africa to develop its own AI capacity. African countries should give clear authorisation for training AI in their copyright laws.
The related issue of whether copyright extends to AI outputs is discussed in a companion brief: Generative AI and sustainable creativity that explains why African countries shouldn’t extend copyright to AI outputs.
Suggested citation
Rens, A., Hlomani, H, & Msipa, S. (2023). Clarifying copyright to enable AI research in Africa. Generative AI and Intellectual Property Brief no. 1. Research ICT Africa. https://researchictafrica.net/publication/ai-and-intellectual-property-brief-1/ |
