The rapidly expanding artificial intelligence industry is facing a major legal and ethical question over whether companies should be allowed to use copyrighted books to train AI models without obtaining permission from authors.
The debate has become one of the most important issues surrounding generative AI because modern systems depend heavily on enormous collections of text and other digital material.
AI companies argue that training models requires access to large amounts of information.
Authors and rights holders, however, increasingly question whether companies should be allowed to commercially benefit from their work without consent or compensation.
How AI training works
Large language models are trained using enormous datasets.
These datasets can contain books, academic papers, websites, articles and other forms of written material.
The systems analyse patterns within that information and use them to generate new responses.
However, the process has raised concerns among writers who argue that their creative work is effectively being used as an input into products that could compete with them.
The issue becomes particularly sensitive when AI companies use copyrighted material to develop commercial systems.
Why authors are concerned
Authors depend on book sales, licensing, speaking engagements and other forms of intellectual-property income.
Generative AI introduces a new form of competition.
An AI system can produce summaries, stories, essays and other written material within seconds.
While AI-generated writing is not necessarily equivalent to professionally produced books, the technology is improving rapidly.
Authors therefore fear that unrestricted access to their work could reduce the economic value of their intellectual property.
The issue is not simply about whether AI systems copy individual books word-for-word.
It is also about whether copyrighted works can legally be used to teach an AI model how to produce language.
The fair-use argument
AI companies have argued in various legal disputes that using copyrighted material for training can fall under doctrines such as fair use in the United States.
The argument is that the technology transforms the original material rather than simply reproducing it.
However, rights holders have challenged that interpretation.
They argue that commercial AI companies should not automatically receive permission to use copyrighted material simply because the resulting system produces something new.
The outcome could have major consequences.
If courts require explicit permission for training data, AI companies could face significant costs.
They might have to negotiate licensing agreements with publishers, authors and other rights holders.
A new market for data licensing
One possible consequence could be the creation of a much larger market for AI training data.
Publishers could license books to AI companies.
News organisations could license archives.
Academic institutions could negotiate access to research.
Individual creators could potentially receive compensation based on how their work is used.
Such a system could establish a more sustainable relationship between AI developers and content creators.
However, determining how compensation should be calculated would be difficult.
Implications for Africa and Nigeria
The issue is also relevant to Nigerian creators.
Nigeria has a growing technology and creative economy.
Authors, journalists, musicians, photographers and other creators are increasingly producing digital content.
As AI adoption expands across Africa, questions about ownership and compensation will become more important.
Nigerian creators may eventually need clearer mechanisms for licensing their work to AI companies.
At the same time, AI developers need access to useful local data.
This could create opportunities for legitimate data partnerships.
The bigger question
The copyright debate ultimately asks a fundamental question: Who owns the information that teaches artificial intelligence?
The answer could determine how the AI industry develops over the next decade.
If creators receive stronger rights, AI companies may face higher development costs.
If courts give companies broad freedom to use copyrighted material, creators may have fewer opportunities to monetise their work.
The eventual balance will likely require a combination of legislation, licensing agreements and court decisions.
For now, the conflict remains unresolved.
But as AI becomes increasingly embedded in education, publishing, business and entertainment, the rules surrounding training data will become just as important as the technology itself.
