Alsuwaidi & Company

Artificial Creativity, Real Copyright Consequences

What happens when a machine creates something that no one can clearly claim to own, using material it may never have been authorised to use?

Artificial intelligence is challenging copyright law at both ends of the creative process: the work it consumes and the content it produces. Between them lies an increasingly complex question of legal responsibility, particularly when AI-generated content crosses borders and enters commercial markets.

Can AI-Generated Works Qualify for Copyright Protection?

Copyright protection has traditionally rested on the principles of authorship and originality, recognising creative expression attributable to a human author. Generative AI challenges this foundation by producing sophisticated literary, artistic and audiovisual content with varying degrees of human involvement, raising fundamental questions about what constitutes authorship and who may claim ownership of the resulting work.

Consider an illustration generated through an AI platform in response to a user’s instructions. Does the act of formulating a prompt constitute sufficient creative contribution to establish authorship? Can the developer of the AI system claim rights over the resulting image merely by virtue of having developed the technology? Or could the illustration remain outside copyright protection altogether, despite its apparent originality and commercial value?

The legal question is not whether AI can produce creative-looking content, but whether the creative expression embodied in that content can be attributed to a legally recognised author.

The distinction between AI-generated and AI-assisted work is essential. Where AI supports identifiable human creative contributions, copyright protection may be available. Conversely, works generated autonomously by AI may fail to satisfy authorship requirements in jurisdictions that require human creativity.

The United States Copyright Office reinforced this distinction  in its report Copyright and Artificial Intelligence, Part 2: Copyrightability (29 January 2025), recognising that AI-assisted works may qualify for copyright protection where sufficient human authorship exists, while prompting alone is insufficient to establish authorship of AI-generated outputs.

In the UAE, Federal Decree-Law No. 38 of 2021 on Copyright and Neighboring Rights defines an author by reference to the person who creates a work and protects various categories of literary, artistic and audiovisual creations. However, it does not expressly address authorship of autonomously generated AI content.

Accordingly, copyright protection for AI-assisted works must be assessed against existing legal principles, considering the extent of human creative involvement and the allocation of economic rights.

The decisive issue is not whether AI was used, but whether protectable human creativity can be identified in the resulting work.

Can Copyrighted Material Lawfully Be Used to Train AI Systems Without Permission?

Behind every generative AI system lies an enormous volume of data, much of it drawn from existing creative work. Books, photographs, music, artwork and software may all contribute to training models capable of producing new content. Yet this process raises a fundamental legal question: does using copyrighted works to train an AI system constitute a legitimate form of technological learning, or an unauthorised exploitation of another’s intellectual property?

The legal controversy surrounding AI training lies not in the ability of technology to analyse existing works, but in the methods through which those works are obtained, reproduced and processed. While copyright does not protect ideas, facts or general concepts, it does protect their original expression and grants rights holders control over specified uses of their works. AI training may involve systematic reproduction, extraction and storage of substantial volumes of copyrighted material, potentially engaging those exclusive rights even where the resulting model does not reproduce the original works verbatim.

Crucially, public availability should not be confused with permission: a work accessible on the internet remains protected by copyright unless its protection has expired, an applicable exception permits the intended use, or the rights holder has authorised it.

The legal treatment of such activities varies considerably across jurisdictions. In the United States, AI training disputes are examined primarily through the fair use doctrine under Section 107 of the US Copyright Act, which requires a case-specific assessment of four factors: the purpose and character of the use, the nature of the copyrighted work, the amount and substantiality of the material used, and the effect on the potential market for the original work.

The distinction between training and the acquisition of training materials is particularly significant. In Bartz v. Anthropic, the US District Court for the Northern District of California held in June 2025 that the use of copyrighted books to train AI models constituted fair use on the facts before it, while rejecting fair-use protection for the creation of a permanent library from pirated copies. The decision illustrates why neither the transformative nature of AI training nor the commercial purpose of an AI system can, in isolation, determine the legality of the underlying use.

The European Union adopts a more specifically regulated approach. Articles 3 and 4 of Directive (EU) 2019/790 on Copyright in the Digital Single Market establish exceptions for text and data mining, distinguishing scientific research undertaken by qualifying institutions from broader mining activities. Article 4 permits certain reproductions and extractions of lawfully accessible works, including for commercial purposes, provided the rights holder has not expressly reserved those uses in an appropriate manner. For publicly available online content, such reservations may be expressed through machine-readable means. The EU framework therefore recognises the potential value of large-scale data analysis while preserving mechanisms through which copyright owners may exercise control over the use of their works.

In the UAE, the legal position must be assessed within the existing copyright framework. Article 7 of Federal Decree-Law No. 38 of 2021 recognises the rights holder’s exclusive economic rights, including reproduction and electronic storage, while Article 22 provides narrowly defined exceptions, some of which are subject to non-commercial limitations. These provisions do not expressly establish a general exception permitting the large-scale use of copyrighted works for commercial AI training. Consequently, the legality of such activities depends on whether the relevant acts fall within the rights holder’s exclusive rights and, if so, whether they are authorised or covered by an applicable statutory exception.

The central issue is therefore not simply whether AI systems may learn from existing works, but whether the methods used to enable that learning respect the legal boundaries of copyright protection.

The challenge lies in distinguishing permissible technological processing from unauthorised exploitation of copyrighted works.

Importantly, lawful training does not necessarily guarantee lawful outputs.

When the Resulting Content Infringes Existing Rights, Who Bears the Liability?

When AI-generated content reproduces protected elements of existing work, determining liability becomes particularly complex. Unlike conventional copyright infringement, several parties may be involved: the AI developer who trains the model, the platform provider who makes it accessible, the user who generates the content, and the business that commercially exploits it. Liability, however, is not automatically shared among them; it depends on each party’s conduct and the applicable legal framework.

Importantly, resemblance alone does not establish infringement. Copyright protects original expression rather than ideas, styles or general concepts. This distinction is reflected in Article 3 of UAE Federal Decree-Law No. 38 of 2021, which excludes abstract ideas and methods from protection while recognising their innovative expression.

Where infringement is established, Article 43 of the UAE Copyright Law provides a basis for rights holders to seek compensation for violations of their rights.

The central challenge is therefore twofold: establishing whether protected expression has been unlawfully reproduced and identifying which party bears legal responsibility for that infringement.

For businesses, the ability to generate content through an AI platform does not guarantee that the resulting material is free from third-party copyright claims.

When AI-Generated Infringement Crosses Borders

Cross-border exploitation introduces an additional layer of legal complexity.

Consider a video game developer outside the UAE using generative AI to create characters or artwork that allegedly reproduce protected elements of an existing game. The resulting product is then distributed internationally, including in the UAE.

Which country’s copyright law applies? Can the rights holder pursue proceedings in the UAE? And should liability rest with the developer, publisher, distributor or another party?

These questions require distinguishing between applicable law, judicial jurisdiction and responsibility for infringement.

Copyright protection is territorial. Conduct considered permissible under one country’s law may not necessarily receive the same treatment elsewhere. Equally, the accessibility of digital content in a particular market does not automatically resolve questions of jurisdiction or liability.

UAE Federal Decree-Law No. 38 of 2021 provides a framework for protecting copyright where relevant rights are infringed within the State. However, cross-border disputes require careful assessment of the infringing acts, the parties involved and the available legal remedies.

For internationally distributed AI-generated content, the location of the developer is only one part of the equation; where and how the content is exploited may be equally significant.

Conclusion

Artificial intelligence has not displaced the fundamental principles of copyright law, but it has challenged their application to increasingly automated and interconnected creative processes.

Questions of authorship, lawful training and infringement liability remain closely linked, particularly where AI-generated content crosses national borders.

For businesses, developers and rights holders, understanding these distinctions is essential to managing intellectual property risks while embracing technological innovation.

The future of copyright in the age of AI will depend not simply on what machines can create, but on how the law protects creativity and allocates responsibility for what they produce.

For advice on copyright protection, AI-related intellectual property risks and cross-border copyright enforcement, please contact the author, Sara Omran, Director, IP Strategy & Practice Development, at s.omran@alsuwaidi.ae