Every day, millions of Afrobeats lyrics are uploaded to platforms like Genius, YouTube, Instagram, and blogs.
Lines from Burna Boy’s ‘Last Last,’ Asake’s ‘Lonely At The Top,’ and Rema’s ‘Calm Down’ sit openly online alongside verses from thousands of independent artists. Now, a growing number of Nigerian artists, producers, and publishers are asking a simple question: when companies like OpenAI train ChatGPT and other large language models, are they quietly copying those same Afrobeats lyrics without permission or payment?
A court ruling halfway across the world has just given the question new urgency.
On November 11, 2025, a Munich regional court ruled that OpenAI’s ChatGPT violated German copyright law by using protected song lyrics without licences from collecting society GEMA.
The court found that the model had memorised entire verses during training and could reproduce them on demand. OpenAI was ordered to pay damages.
Read also: Davido’s ‘Coming to America 2’ sync deal highlights Afrobeats’ global business potential
The German decision is the clearest signal yet that training AI on copyrighted lyrics without permission is illegal in many jurisdictions. For Nigeria’s Afrobeats industry – Africa’s biggest music export and one of the fastest-growing genres worldwide – the implications are direct: if lyrics posted online count as ‘publicly available data,’ then millions of Nigerian verses may already be inside the world’s most popular AI systems, with no licence and no money reaching the creators.
The decision has put a spotlight on Nigeria’s booming Afrobeats scene. Artists, producers, publishers, and lawyers are now asking the same question: are the world’s biggest AI models quietly training on millions of Nigerian lyrics scraped from Genius, YouTube descriptions, blogs, and social media – all without a single licence or payment reaching the creators?
No Nigerian-specific evidence, but no transparency either
OpenAI has never published a detailed list of training data for GPT-4, GPT-4o, or the models behind ChatGPT. The company states that it uses ‘publicly available’ internet data, which almost certainly includes song lyrics posted online. Afrobeats lyrics – from Burna Boy, Wizkid, Asake, Rema, and thousands of independent artists – are widely shared on free platforms, making them easy to collect.
In practice, this means that when a user asks ChatGPT to “write a verse in the style of Wizkid” or “create Afrobeats lyrics about hustle,” the model may be drawing on patterns it learned from real Nigerian songs during training.
Whether it can reproduce exact lines depends on how strongly those lines were memorised, but the German ruling confirmed that memorisation and reproduction do happen.
Nigeria’s Copyright Law has no AI rules yet
Adekunle Shorinola, managing director of Lagos-based music publishing company Durozy, said Nigeria’s Copyright Act of 2023 contains no provisions that address the training of AI models.
“The law has not caught up with the technology,” Shorinola told reporters. “Right now, if an AI company scrapes lyrics from the internet and uses them to train a model, there is no clear section of Nigerian law that stops them or forces them to pay.”
He added that even basic royalty collection for human-made music remains weak in Nigeria because collecting societies and some industry leaders do not fully understand digital income streams. “If we are still struggling to collect ordinary performance and mechanical royalties, how can we police AI training?” he asked.
Major labels are already signing AI deals
On November 20, 2025, Los Angeles-based music-tech company KLAY Vision announced licensing agreements with Universal Music Group, Sony Music Entertainment, Warner Music Group, and their publishing arms. These deals allow KLAY to legally train its AI tools on the entire catalogues of the three biggest music companies in exchange for agreed fees and protections for artists.
Akachi Igboko, an A&R executive at one of Nigeria’s largest labels, said that such deals only cover artists signed to those majors. The vast majority of Afrobeats creators – independent artists, upcoming producers, and small labels – are not part of any similar agreement.
Can AI infringement be proven today?
Both experts agree that proving infringement is difficult but not impossible.
Igboko said that because AI is here to democratise music creation, proving infringement is extremely difficult. It may only be possible if an infringement is detected from an AI-generated track that is gaining traction, allowing a creator to then demonstrate the recreation of a melody or specific lines of the original song.
“An artist would need to show screenshots or recordings of the AI outputting their exact lyrics or melody, then hire a musicologist and a lawyer,” Igboko said.
Read also: The Afrobeats Film Movement takes center stage at AFRIFF
Practical steps Nigerian creators can take now
While waiting for laws to change, industry players at music festivals and workshops suggest that artists should register their works with the Nigerian Copyright Commission and with international collecting societies (PRS for music that are already fighting AI cases).
Igboko said artists and labels should consider ‘ethical training’ deals where some platforms now let artists voluntarily upload their catalogue to train licensed AI voices or writing styles and earn royalties when the model is used.
Shorinola noted that there should be a push for the Copyright Society of Nigeria (COSON) and lawmakers to add AI-training clauses to the Copyright Act to meet new international standards.
The road ahead
The German precedent, combined with ongoing lawsuits in the United States against OpenAI, Anthropic, and Stability AI, shows that courts are willing to treat training on copyrighted works as infringement when no licence exists.
For Nigeria’s music industry – now the second-largest in Africa by recorded music revenue and a global streaming powerhouse – the stakes are high. Without swift updates to the law and stronger collective licensing, the beats, slang, and stories that define Afrobeats could continue powering foreign AI systems while the originators receive nothing.
