Imagine that you spend years writing a novel or painting a painting, and suddenly you discover that an artificial intelligence has used your work—without asking you or paying you—to learn and generate similar content.
The explosion of generative AI (such as ChatGPT, MidJourney or Gemini) has reopened an uncomfortable debate: what is worth more, the right of authors to protect their work or the freedom of companies to train their algorithms with any available data?
Giants like OpenAI, Google and Meta insist that “fair use” protects their practices, while artists, writers and musicians denounce plagiarism. Even figures like Elon Musk and Jack Dorsey are calling for abolishing copyright laws, arguing that they slow down innovation.
The tension between human creation and algorithmic generation is evident. In this context, AI and the current regulatory framework face the challenge of defining whether intellectual property laws should be sacrificed or adapted to technological progress.
The conflict explained
Behind every AI model is an insatiable appetite for data: millions of books, songs, photos, and articles that fuel your learning. The problem arises when these contents are protected by copyright.
Companies like OpenAI, Google, and Meta have built their systems using other people’s works without obtaining permission or compensating the original creators.
The technology companies defend their position with two key arguments: first, they allege that their use falls under “fair use”, a legal figure that in the US allows the limited use of protected material without a license for purposes such as research or parody.
They also warn that, if they cannot freely access this data, they will lose the technological race to countries like China, where restrictions are fewer. However, artists, writers and journalists see this as exploitation disguised as innovation.
Cases such as that of Sarah Andersen, illustrator plaintiff in a historic trial against Stability AI, show the discomfort: her artistic style was replicated by AI without her consent, diluting her original work in a sea of algorithmic imitations.
Some specific cases
The controversy has jumped from theoretical debates to the courts with lawsuits that could set precedents.
The New York Times sued OpenAI and Microsoft for using millions of their articles to train ChatGPT without permission or compensation, claiming that the model is now a competing source of information.
In the art world, Kelly McKernan, Sarah Andersen, and Karla Ortiz led a class-action lawsuit against Stability AI, MidJourney, and DeviantArt, accusing them of using their protected artwork to train image generators that replicate their styles.
In music, Universal Music Group has blocked attempts to clone voices of artists like Drake or The Weeknd using AI, while platforms like GitHub faced lawsuits over its Copilot tool, which suggested code based on private repositories.
These cases reveal a pattern: the AI industry operates under the premise of “asking for forgiveness instead of permission,” while creators demand that the technology not override their rights.**
The legal vacuum: a system that failed to anticipate AI
Copyright laws were created in an analog era, with physical copies in mind, not algorithms that learn from patterns. This gap has created a gray area where AI companies operate with relative impunity.
The core of the conflict is that current legislation does not clearly define whether model training constitutes a violation of rights or falls under exceptions such as “fair use.”
As the EU moves forward with its AI Act—which requires transparency in training data—and Japan explicitly allows the use of copyrighted material for AI, the US remains in a dangerous ambiguity.
Federal courts have issued conflicting rulings: some judges see the process as “transformative” (favoring AI), while others insist that simply accessing protected works without a license is already illegal.**
This limbo benefits technology companies, but leaves creators at a disadvantage. Without urgent legal reform, justice will continue to arrive late to each controversy.
Consequences: an ecosystem at risk
The conflict between AI and copyright is creating a domino effect with profound ramifications.
Creators
For creators, the most immediate consequence is the devaluation of their work: when an AI can generate similar content in seconds using their works as a basis, they lose control over their art and their ability to monetize it.
Platforms are already flooded with books, illustrations and music generated by AI that compete directly with human works, but without investing in their creation.
Companies
For companies, the risk is a crisis of legitimacy. Million-dollar lawsuits and protests by artists are damaging their reputation, while legal uncertainty slows down investments.
In the medium term, it could lead them to self-regulate – as OpenAI did by signing agreements with some media – but only with those creators with sufficient negotiating power.
Society
And for society, the danger is a homogenized culture: if AI only recycles what exists without permission, real innovation is eroded. Without incentives to create, what will feed future generations of AI? The true cost could be human creativity itself.
Possible solutions: Towards a sustainable balance
The conflict requires creative solutions that reconcile innovation and rights. One option is collective licenses, such as those managed by author societies, where companies would pay fees to funds that compensate creators.
Another path is radical transparency: for AI developments to detail what data they used, allowing authors to opt out of their works (as proposed by the EU AI Act).
The use of ethical data banks could also be promoted, with works in the public domain or voluntarily transferred. Startups like Stable Audio are already exploring this model. The key is that technology must advance without extinguishing the sources that feed it.
What consensus can be reached in this regard?
Balancing AI and copyright does not require abolishing standards, but adapting them. Compulsory licenses with fair payment could be applied, similar to those in the music industry, where companies pay to use protected works at standardized rates.
In parallel, algorithmic transparency—that developers reveal what data they use—should be required, giving creators the right to exclude their works if they wish.
This model would satisfy both parties: technology companies would access the data necessary to innovate, while the authors would receive recognition and remuneration.
The key is to regulate without strangling: policies that promote ethical AI without extinguishing the incentives to create. Consensus is possible, but it requires a willingness to cooperate, something that until now has been conspicuous by its absence.
This post is also available in: