When we look at an AI-generated image, such as Midjourney or DALL-E, we are faced with a more complex question than it seems: who owns the intellectual property rights to this content?
The question is not only theoretical but has concrete implications for companies, creatives, developers, and visual communication professionals. In a context where AI is increasingly integrated into production flows, clarifying who owns the copyright becomes a priority, both legally and ethically.
We explored this issue in the last installment of Neanche un minuto di AI. In this article, we focus on the practical and regulatory aspects related to the intellectual property of AI-generated content.
What is intellectual property (and why is it an issue in AI)?
Intellectual property protects the creations of the human mind: works of art, texts, music, inventions. In the case of AI-generated images, the main node concerns copyright, which gives the author exclusive control over the use and dissemination of the work.
However, according to the law, the author must be a natural person. Artificial intelligence, not being a legal entity, cannot hold rights. This creates a regulatory vacuum that calls into question the ownership of works created by AI models.
Can AI be considered an author?
No. According to Italian, European and international regulations, only a human being can be recognized as the author. If the image is generated entirely by an AI, without relevant human creative intervention, it cannot therefore be protected by copyright.
This principle has been confirmed by bodies such as the U.S. Copyright Office, which has refused to register works produced solely by Artificial Intelligence. In summary: without a human author, there can be no copyright.
What about the user who writes the prompt?
Here the field gets slippery. Writing a prompt – that is, the instruction that guides the AI in generating the image – can be a very simple action or more detailed, requiring artistic, semantic and technical skills.
Some experts argue that if the user’s contribution is sufficiently creative and original, it can be justified for the user to be given the rights. At present, however, there is no objective criterion.
In Italy, as in other EU countries, there is a tendency to evaluate on a case-by-case basis, looking for the so-called “relevant creative contribution.” If the prompt is trivial or automated, rights are generally not awarded to the user.
Can the company developing the model claim rights?
Companies developing AI models can determine, through terms of service, who can use the generated images and in what ways. However, these are contractual licenses, not actual copyrights.
Some platforms grant full freedom of commercial use, while others impose more stringent limitations. In this scenario, the company can maintain indirect control over the images while not being legally recognized as the author, an approach that relies more on software licensing logics than classic artistic intellectual property.
What are the risks to companies?
For companies using generative AI to create visual content, the risks are not theoretical but concrete:
- Ambiguity over rights ownership: in the absence of clear regulations, it is difficult to protect their creations from copying, improper reuse, or legal disputes.
- Compliance problems: generated content could be based on datasets that include copyrighted works, with the risk of infringement against third-party authors.
- Reputational damage: lack of transparency in the use of AI can undermine user and public trust.
This is why it is crucial to define clear internal policies, adopt tools for tracking the origin of images, and train business teams on the limitations and potential of these technologies.
An evolving regulatory landscape
In 2024, the European Parliament passed the AI Act, the first organic legislation on Artificial Intelligence. While it does not directly address the issue of copyright, it paves the way for a possible legal framework for AI-generated content, promoting fundamental principles such as transparency, traceability and accountability in the use of models.
In parallel, the World Intellectual Property Organization (WIPO) and authorities in the United States, United Kingdom, and Japan are also working on reforms to adapt the copyright system to the AI era. There are still many knots to untangle, but the debate is more alive than ever.
Plagiarism and “stylistic influences”: where does the line cross?
Among the most sensitive issues related to intellectual property is that of plagiarism. If an AI-generated image replicates the style of a recognizable author-such as Klimt, Van Gogh, or Banksy-is it merely an inspiration or an unauthorized copy?
Technically speaking, many AI models are trained on huge datasets that include millions of images, some of which may be copyrighted.
This has already led to lawsuits from artists, photographers and illustrators, who accuse tech companies of outright “digital plunder” of their works. However, proving plagiarism in court is not so easy, precisely because the output generated by these systems is the result of probabilistic and transformative processes that are difficult to trace back to direct copying.
In the absence of firm rules, the future of intellectual property applied to AI will play out on two main fronts:
- New ad hoc laws, recognizing – or excluding – specific rights for AI-generated content, perhaps introducing criteria for “co-creation” between humans and machines.
- Ethical and contractual standards, promoted by companies and institutions to ensure transparency, author protection and accountability in the use of generative AI.
While waiting for a more defined legal framework, those who use these technologies-including companies-must move with awareness, balance and care to avoid incurring legal, reputational or ethical risks.