{"id":33469,"date":"2026-06-24T13:59:47","date_gmt":"2026-06-24T13:59:47","guid":{"rendered":"https:\/\/wedodis.co.za\/news\/?p=33469"},"modified":"2026-06-24T13:59:55","modified_gmt":"2026-06-24T13:59:55","slug":"sol-phenduka-reveals-sa-artists-featured-in-ai-training-data","status":"publish","type":"post","link":"https:\/\/wedodis.co.za\/news\/sol-phenduka-reveals-sa-artists-featured-in-ai-training-data\/","title":{"rendered":"Sol Phenduka Reveals SA Artists Featured in AI Training Data"},"content":{"rendered":"
\n

Prominent broadcaster Sol Phenduka<\/a> responded after discovering that South African music contributed to the training of AI models. This revelation follows an article in The Atlantic<\/i> that unveiled an accessible database revealing which artists and tracks have been incorporated into the training of Suno and Udio AI models.<\/p>\n

This issue affects countless musicians, regardless of their fame. The news has led to significant outrage from artists regarding the unauthorized use of their music for training AI systems. Sol Phenduka, notably known for creating a song that won the 2024 Song of the Year on Lesedi FM, expressed his feelings about his music being used for AI training.<\/p>\n

Sol Phenduka responds to inclusion of SA artists in AI training datasets<\/h2>\n

On his official X (Twitter) account on Monday, June 22, 2026, Sol Phenduka posted a chart displaying South African musicians whose work has been utilized to train AI systems. In his post, the former Kaya FM presenter indicated his disappointment that both his work and that of various other South African artists were among those listed in the datasets. The post stated:<\/p>\n

\n

\u201cSome South African artists found in AI training datasets💔\u201d<\/p>\n<\/blockquote>\n

A Cybersec audit of the AI training data identified Kabza De Small as the most affected South African artist, followed by DJ Maphorisa. Spirit Praise ranks third, and Nasty C is noted as the highest-placed hip-hop artist from South Africa. The music was incorporated into AI models designed to create similar songs based on simple prompts from users.<\/p>\n

See the full infographic below<\/a>:<\/p>\n

The response to the inclusion of musicians in AI training datasets<\/h2>\n

Sol Phenduka\u2019s post reignited discussions around issues of copyright, consent, and artistic ownership in the context of artificial intelligence.<\/p>\n

Responses included:<\/p>\n

@itsLord_za stated:<\/p>\n

\n

\u201cSouth African artists built culture. If AI is learning from their work, the conversation about consent, credit and compensation cannot be ignored.\u201d<\/p>\n<\/blockquote>\n

@NkunziWe elaborated:<\/p>\n

\n

\u201cIn simpler terms, AI systems learn by analysing huge amounts of information called training data. If an artist\u2019s work is in that data, the AI may have \u201cstudied\u201d it during training. This does not mean the AI copies the artist\u2019s work, but it indicates the AI learned patterns, styles, or information from it.\u201d<\/p>\n<\/blockquote>\n

@MunchoB shared:<\/p>\n

\n

\u201cWhy is it heartbreaking? When Fruity Loops dropped, people who were in instrumental bands downplayed it, but look where we are with Fruity Loops.\u201d<\/p>\n<\/blockquote>\n

@advmofokeng1 argued:<\/p>\n

\n

\u201cIt is very simple, they just sue. I don’t see what is difficult there. I am sure there is already an ongoing case in the US. The authors have already had theirs settled. Bo bra @ZakesMda got some payment for similar violations.\u201d<\/p>\n<\/blockquote>\n<\/div>\n\n\n

\n

Some South African artists found in Ai training datasets .

\ud83d\udc94 pic.twitter.com\/vCnO6oat6B<\/a><\/p>— Thandubuntu (@Solphendukaa) June 22, 2026<\/a><\/blockquote>