A recent audit of large-scale artificial intelligence training corpora has identified hundreds of South African artists whose musical works are included in datasets used to train generative AI systems. The findings, published by Cybersec Clinique on 21 June 2026, were made possible through a searchable database released by The Atlantic, which allows for the tracking of audio tracks used across 21 million entries.
Prominent figures in the local music industry, including Kabza De Small, DJ Maphorisa, Spirit of Praise, Nasty C, Thomas Chauke, Brenda Fassie, and Selaelo Selota, were among those found within the training materials. The discovery has prompted significant conversation regarding the ethics of using cultural output to build commercial AI models without explicit permission or compensation.
Concerns Over Cultural Data Ownership
Speaking on the broader implications of the findings, Doreen Mokoena, founder and CEO of Cybersec Clinique, stated that the situation represents a systemic challenge for local creators. The biggest export South Africa never authorised may be its culture,” Mokoena noted following the release of the data.
As artists, rights holders, and policymakers begin to examine the scale of this practice, it may prove to be one of the most significant data laundering controversies the global music industry has ever faced.
Legal And Technical Implications
While the presence of a song in a training dataset does not serve as definitive proof of copyright infringement, industry experts maintain that it constitutes a major development in the ongoing debate over intellectual property in the digital age. The current South African Copyright Act of 1978 does not explicitly address the ingestion of creative works by machine learning models, leaving many artists and industry bodies to call for urgent legislative reform.
Researchers caution that verifying whether an AI model specifically mimics an individual artist’s unique style requires further technical analysis beyond simply identifying the presence of their work in a database. Community debate has intensified since social media personality Sol Phenduka shared the findings on 22 June, highlighting the growing apprehension among performers regarding the future of their revenue streams and creative autonomy.
Further developments are expected as local unions and music associations review the findings to determine potential collective action to protect the rights of South African musicians.