Artificial Intelligence

AI, copyright and authorship: who really owns a work created with artificial intelligence?

The debate over intellectual property, creativity and authorship in the age of generative AI. Real cases, legal dilemmas and a question that redefines human creativity.

Jorge Louis Fernández Heredia · 14 min read · May 2026

There was a time when authorship seemed relatively straightforward. If someone built a house, there were blueprints, permits, contracts and an architect's signature. If someone painted a work, it was possible to study the brushstrokes, technique and pigments to determine whether it truly belonged to Picasso, Dalí or Van Gogh. Intellectual property was tied to something deeply human: manual skill, creative intent and the physical evidence of a process.

Then came the internet. And with the internet came a new problem: digital content could be copied infinitely, modified and redistributed in seconds. A photograph published on a website could end up on thousands of different sites. An article could be copied, summarized, translated or slightly modified until it was unrecognizable compared to the original. The first discussions about digital copyright began there.

But what we are experiencing today with artificial intelligence is not simply an evolution of that problem. It is something far more profound. Because for the first time in modern history we are entering a scenario where creation itself is becoming blurred. And that completely changes our understanding of authorship.

The big question: who owns a work created by AI?

Consider the following scenario:

"Generate a futuristic city in the rain with a cyberpunk aesthetic."

The AI produces four images. The person is not satisfied. They iterate. Modify words. Request lighting changes. Adjust colors. Reframe the shot. Discard versions. After 50 iterations a spectacular image appears. They publish it.

Then the question arises:

  • Is the author the person who wrote the prompt?
  • Is the AI the author?
  • Is the company that trained the model the author?
  • Are the millions of artists whose works were used to train that AI indirect authors?
  • Or does the work simply have no clear owner?

This is where current laws begin to strain. Because virtually all modern copyright legislation was designed under a fundamental premise: every work has an identifiable human creator. AI partially breaks that logic. But that does not necessarily mean the human disappears from the creative process.

In fact, the opposite is likely to happen.

The mistake of thinking AI replaces human creativity

There is a very popular narrative suggesting that AI will eventually create better than humans. But perhaps that is not the right discussion.

AI has no consciousness. No intention. No emotions. No experiences. No real human context. It operates on probabilities. And that difference is enormous.

When an AI generates an image, a text or a song, it is not "imagining" something the way a person would. It is statistically calculating which next word, pixel or note is most likely based on patterns learned from enormous quantities of prior human content.

AI does not create from human experience; it creates from mathematical probability.

And that makes the conversation far more interesting.

The infinite monkey and artificial creativity

There is a famous philosophical and mathematical thought experiment known as the infinite monkey theorem. The idea is simple: if a monkey presses keys at random for an infinite amount of time, it could eventually write all of Shakespeare's works. Mathematically, it is possible. But that does not make the monkey Shakespeare. Nor does it transform the result into a work born from conscious artistic intent.

Human creativity is not just the final result

It also involves intention, context, experience, emotion, meaning, purpose and interpretation.

And yet, AI forces us to face an uncomfortable question:

How much of what we called human creativity was truly intent… and how much was iteration, error, probability and accidental discovery?

Because even humans often create this way. We try. We fail. We iterate. We discover happy accidents. Many artistic movements were born from mistakes. Many discoveries happened accidentally. Many masterpieces emerged from reinterpretations of earlier ideas.

AI does not necessarily destroy human creativity. But it does expose something uncomfortable: perhaps creativity was always more emergent and less "magical" than we wanted to admit.

What remains exclusively human?

Perhaps the answer does not lie in generating content. Perhaps it lies in recognizing meaning.

Because even if an AI or a probabilistic system could generate something extraordinary, only a human would be able to:

  • be moved by it,
  • interpret it,
  • give it context,
  • find purpose in it,
  • attribute cultural value to it,
  • connect that work to a human experience.

The infinite monkey might accidentally write the Bible. But only a human could recognize that it was the Bible. And that difference will likely remain fundamental for a very long time.

From artist to creative director

Perhaps the real shift is not that AI replaces artists. Perhaps the shift is that it redefines where creative value lies.

For centuries value was deeply tied to manual execution: painting well, writing well, designing well, performing well. But when an AI can generate thousands of variations in seconds, value begins to shift toward other abilities:

  • having vision,
  • selecting,
  • directing,
  • curating,
  • deciding,
  • building identity,
  • contributing meaning.

The human stops being merely the one who executes. And begins to look more like a creative director, an editor, a curator, or even a conductor.

The orchestra metaphor

Imagine a symphony orchestra. There is a composer who wrote the original piece. There is a conductor who interprets that work through their own vision. And there are dozens of musicians who perform each part.

But something interesting happens: sometimes an exceptional musician adds an unexpected nuance. A different interpretation. A new emotion. A small improvised detail. The conductor hears it and decides to incorporate it permanently into future performances.

Then the conflict emerges:

  • Does the work still belong solely to the composer?
  • Did the musician contribute their own creativity?
  • Does the full performance become a distinct work?

AI works in a similar way. The human directs. The AI interprets probabilistically. But during that interpretation, unexpected results appear that even the user had not imagined.

The best part of the result does not come exactly from the initial human idea… but from what emerges during the interaction between human and model.

That is where the traditional concept of authorship begins to fracture.

The invisible problem: works used to train AI

AI models did not emerge from nothing. They were trained on enormous quantities of human content: photographs, illustrations, films, articles, music, code, voices, complete books. In many cases, without explicit authorization from their creators.

And that opened multi-million dollar lawsuits around the world.

Artists began discovering that generative models could replicate styles extremely similar to their own. Writers discovered that their books had been used as training material. Media organizations began to notice that AI-generated responses drastically reduced traffic to their sites.

At that point the discussion stops being artistic. It becomes economic.

Real cases and references

The New York Times v. OpenAI and Microsoft

In December 2023, The New York Times sued OpenAI and Microsoft, accusing them of using millions of journalistic articles to train AI models without authorization.
Source: The GuardianNew York Times sues OpenAI and Microsoft over AI training data

The lawsuit argues that OpenAI and Microsoft used Times content to build products that directly compete with the newspaper, without compensating the original creators of that content.
Source: Harvard Law ReviewNYT v. OpenAI: The Times's About-Face

Getty Images v. Stability AI

Getty Images took legal action against Stability AI, accusing the company of using millions of copyright-protected images to train the Stable Diffusion model without authorization.
Source: BakerLawGetty Images v. Stability AI

A UK court also admitted the case, making it an important precedent for visual content protection in Europe.
Source: UK JudiciaryGetty Images and others v. Stability AI

Media organizations and AI-powered automatic responses

Various international and Latin American media organizations have begun to question the economic impact of AI-generated automatic responses inside search engines. The central concern is that users obtain part of the information without visiting the original site, reducing traffic and ad revenue for those who produce the content.
Source: OpenAIResponse to NYT data demands
Source: Columbia Law ReviewNYT v. OpenAI and Microsoft

Research on memorization in AI models

Recent academic studies have investigated how advanced models can reproduce portions of copyright-protected works under certain conditions, adding a technical dimension to the legal debate.
Source: arXivMemorization in Large Language Models (2024)
Source: arXivCopyright and Large Language Models (2024)

Will everything belong to everyone?

Probably not. But it does seem that we are heading toward a scenario where:

  • authorship will be more shared,
  • more ambiguous,
  • more collaborative,
  • and much harder to prove.

AI does not eliminate the concept of intellectual property. What it does is weaken its traditional boundaries. Laws will evolve. Courts will establish precedents. And during that process there will be real tension between those who create original content, those who develop models, and those who use them.

The final question

Perhaps the most important question is not:

"Can AI create?"

But another, far more profound one:

"What part of human creativity was always exclusively human… and what part was emergent discovery?"

And depending on how we answer that, we will redefine the concept of authorship for the generations to come.

That redefinition is not abstract or distant. In some countries it already has a name, a face, and a court case. Chile is no exception.

The conflict already happening in Chile

The debate around artificial intelligence and intellectual property has already begun to be felt in Chile.

One of the most talked-about recent cases involves television channels and media organizations questioning how AI-powered automatic responses inside search engines could directly affect their advertising revenue.

The concern is relatively straightforward:

  1. Media outlets invest human and financial resources to produce content.
  2. They publish news, reports and analysis.
  3. They depend on web traffic and digital advertising.
  4. But search engines are now beginning to respond automatically using generative AI.
  5. Users get a significant part of the information without visiting the original site.

The potential result: fewer visits, fewer ad impressions, lower monetization and loss of value for those who produce the original content.

Although this debate is still evolving legally in Chile, it reflects a global phenomenon: the tension between AI platforms that synthesize information and media organizations that depend economically on user attention.
Source: BioBioChileTV channels file lawsuit against Google for abuse of dominant position
Source: La TerceraChilean TV channels sue Google at TDLC for abuse of dominant position
Source: Diario FinancieroSix Chilean TV channels sue Google for anticompetitive practices

Verifiable references:
The Guardian — NYT vs. OpenAI (Dec. 2023) · Harvard Law Review — NYT v. OpenAI analysis · BakerLaw — Getty vs. Stability AI · UK Judiciary — Getty UK judgment · arXiv — Memorization in LLMs · BioBioChile — TV channels vs. Google (Chile) · La Tercera — Chilean TV channels sue Google TDLC · Diario Financiero — Six TV channels vs. Google (Chile)

Share this article

Frequently asked questions

It depends on the country and the level of human participation in the creative process. In the United States, the Copyright Office has indicated that it only registers works with sufficient human authorship. If the human simply wrote a basic prompt and accepted the first result without significant modification, the work likely does not qualify for automatic protection. If there was an iterative process, meaningful creative decisions and substantial modifications, there is a greater possibility that part of the result may be protectable. Legislation on this topic is still evolving in most countries.

It depends on the terms of service of the platform you used and local legislation. Many generative AI platforms state in their terms that the user is responsible for the content they generate and may use it commercially, subject to their restrictions. However, there is also the risk that the model was trained on protected content and that the generated work resembles an original protected work too closely. It is always advisable to review the platform's terms and, for high-stakes uses, to consult an intellectual property specialist.

These are two of the most significant legal cases on AI and copyright. The New York Times sued OpenAI and Microsoft in December 2023, arguing that they used millions of the newspaper's articles to train models without authorization. Getty Images sued Stability AI for using millions of protected images to train Stable Diffusion. Both cases seek to establish whether using protected content to train AI constitutes copyright infringement, and could set important precedents for the entire industry.

In some cases, yes. Academic research has documented that certain large language models can reproduce fragments of works they were trained on, especially when explicitly prompted. This phenomenon is called "memorization" and is one of the most important technical arguments in lawsuits against AI companies. For visual content, it has also been documented that some models can generate images very similar to the style of specific artists who were part of the training data.

Authorship in the age of AI will likely not disappear, but it will become more complex, more collective and harder to define. In the future, there may be legal frameworks that recognize different levels of creative contribution: the model developer's, the user who directs the process, and the original creators whose content was used for training. It is also likely that the value of human authorship will progressively shift from technical execution toward creative direction, curation, intent and meaning.