Generative AI is creating new opportunities for businesses but also exposing a critical legal blind spot. Larissa Maria Cenan, Head of Legal, Ethics & Compliance Officer at Capgemini in Romania examines why ownership of AI-generated content is only part of the challenge, and why contracts must evolve to keep pace with the technology.
When I chosed the topic of my LL.M. dissertation,“Generative Artificial Intelligence: Navigating the Intellectual Property System,” my interest was not purely academic, I must admit. It was shaped by the professional environment in which I work, one where artificial intelligence is increasingly embedded in the way technology-driven services are designed, developed and delivered. As AI tools became part of daily workflows, questions naturally started to arise around the legal status of AI-generated outputs, the protection of intellectual property, and the extent to which existing legal frameworks are prepared to address a new form of technology-assisted creation.
Supporting innovation in a highly technical environment requires more than understanding the technology itself. It requires understanding the legal consequences of using that technology. For me, this was the starting point of the research: the recognition that generative AI is not merely a tool for efficiency, but a legal disruptor that challenges some of the most traditional concepts of intellectual property law, authorship, originality, ownership, liability and exploitation rights.
The central question is deceptively simple: who owns AI-generated content?
The answer, however, is far less straightforward.
Generative AI and the Challenge to Traditional Copyright Concepts
Copyright law has historically been built around human creativity. It protects the expression of ideas by an author and is premised on the assumption that creative works originate from human intellectual effort. As explored in my research, intellectual property law protects intangible assets resulting from original creative thought, including written works, artistic creations, software, methodologies, designs and other creations of the mind.
Generative AI challenges this foundation. Unlike traditional software, generative AI systems do not merely execute predefined instructions. They can generate text, images, music, code and other forms of content based on patterns learned from vast datasets. In this sense, generative AI goes beyond the analysis of data and enters the field of content creation.
This technological capability raises a fundamental legal issue: if a work is generated by an AI system, can it be protected by copyright, and if so, who should be considered the author?
Under Romanian copyright law, the author is expressly defined as the natural person or persons who created the work. This human-centric approach is consistent with the traditional understanding of copyright, where originality is linked to the personal intellectual contribution of a human creator.
The same difficulty arises at a broader legal and doctrinal level. My research highlighted that copyright law remains largely anthropocentric: it searches for the author’s personal imprint in order to establish the originality necessary for protection. A machine, however advanced, does not have personality, legal capacity, moral interests or human creative intention.
This does not mean that AI-generated outputs are irrelevant from an IP perspective. On the contrary, they may have significant economic and commercial value. But their legal qualification remains uncertain.
The Illusion of Automatic Ownership
One of the most important practical risks arising from generative AI is the assumption that whoever prompts the system automatically owns the output. This assumption may be commercially convenient, but it is legally fragile.
If the output does not qualify as a copyright-protected work because it lacks sufficient human creative contribution, there may be no copyright to own. If the output incorporates or closely resembles protected third-party works, ownership becomes even more complex. If the AI tool’s contractual terms reserve certain rights to the provider or restrict commercial use, the user may not have the rights they believe they have.
This is why the ownership question cannot be answered only by looking at copyright statutes. It must also be addressed contractually.
The real legal blind spot is not only whether AI-generated content is protected by copyright. The blind spot is that many contracts still treat AI-generated outputs as ordinary deliverables, without addressing the specific legal risks created by the use of generative AI.
Traditional IP clauses usually state that all intellectual property rights in the deliverables are assigned to the client or remain with the supplier, depending on the commercial model. In AI-related projects, this wording may be insufficient. If there is uncertainty as to whether copyright exists in the first place, a standard IP assignment clause may not provide the level of legal certainty the parties expect.
“Contracts must move from generic ownership language to a more operational allocation of rights, responsibilities and risk.”
Human Authorship, AI Assistance and the Role of Creative Control
A distinction must be made between AI-generated and AI-assisted content.
Where AI is used merely as a tool to support a human creative process , for example, to suggest wording, structure, alternatives or technical options ,the final output may still reflect sufficient human intellectual contribution. In such cases, the human author’s creative choices, selection, arrangement, editing and judgment remain legally relevant.
By contrast, where the AI system autonomously generates the output with minimal human intervention, the argument for copyright protection becomes weaker. As discussed in my dissertation, the key difficulty lies in the first condition for copyright protection: the existence of creative activity attributable to an author. While the requirement that a work should have a concrete form of expression may be easier to satisfy, the requirement of human creative activity remains problematic in the case of AI-generated works.
This distinction is essential for contracts. Parties should not refer generically to “AI content” without clarifying whether the content is generated autonomously, created with human assistance, or merely supported by AI tools. The degree of human involvement may influence whether rights exist, who can claim them, and what warranties can reasonably be provided.
The Contractual Blind Spot
The title of this article refers to a “contractual blind spot” because the legal uncertainty surrounding AI-generated content is often not properly reflected in commercial agreements.
Contracts involving AI-generated or AI-assisted content should no longer rely exclusively on standard IP ownership clauses. They should address, at a minimum:
In other words, contracts must move from generic ownership language to a more operational allocation of rights, responsibilities and risk.
The European Union has taken a leading role in regulating artificial intelligence through Regulation (EU) 2024/1689, the AI Act. This Regulation as a turning point in the regulation and use of AI technologies, with implications for data protection, ethics, fundamental rights, cybersecurity and trust in AI adoption.
Although the AI Act is not a copyright regulation, it is highly relevant to the broader governance of AI systems. It introduces a regulatory framework designed to ensure that AI systems are safe, transparent and aligned with fundamental rights.
Why Legal Teams Should Reframe the Question
The question “who owns AI-generated content?” remains important, but it should not be the only question.
A more complete legal assessment should also ask:
Can the output be protected? Was there sufficient human contribution? Were the inputs lawfully used? Are third-party rights implicated? Do the tool’s terms permit the intended use? Can the output be commercialised? Is the output subject to confidentiality, data protection or regulatory restrictions? Has the risk been allocated contractually?
This broader approach is particularly important because AI-generated content may be fast, scalable and cost-efficient, but it may also be legally uncertain. The speed of technological adoption can easily outpace the maturity of legal controls.
Conclusion: Ownership Is Only Part of the Answer
Generative AI is transforming the way content is created, reviewed, delivered and commercialised. It is also forcing lawyers to revisit some of the most foundational assumptions of intellectual property law.
The classical copyright framework was not designed for autonomous machine-generated outputs. It was designed around human creativity, human authorship and human personality. Collaboration between legal systems, regulators, academia and industry will be essential to ensure that generative AI is used responsibly and that existing legal frameworks evolve in a coherent manner.
In my view, until the law becomes clearer, contracts will play a decisive role.
The real risk is not only that no one knows who owns AI-generated content. The greater risk is that parties assume ownership has been settled when, in reality, the contract does not address the specific legal consequences of using AI at all.
That is the contractual blind spot.
Be part of a growing global community committed to advancing in-house legal leadership.
As AI transforms contract review and legal workflows, Mauro Ribeiro, Head of Legal and Commercial Management at Huawei, explores the part of the General Counsel’s...
Markus Warmholz, Head of Corporate & International Legal Affairs and Legal Operations at Hartmann Group in Germany, highlights overlooked legal pitfalls that can derail even...
