A technology company releases a suite of creative assets – product images, marketing copy, and software modules – each produced entirely by an AI system without direct human authorship. The business assumes it owns all resulting intellectual property. Then a licensing dispute arises, a competitor copies the outputs, and a question surfaces that Irish law has not yet fully answered: who, if anyone, holds enforceable rights in those works?
AI-generated works occupy an uncertain position under Irish intellectual property legislation, which was designed for human authors and has not been formally updated to address autonomous machine creation. Irish copyright law contains a specific provision for computer-generated works. assigning authorship to the person who makes the necessary arrangements. but courts have not yet tested its boundaries in the context of modern AI systems. Businesses operating in Ireland that rely on AI-generated content, code, or design face genuine exposure: rights they believe they hold may be legally fragile or, in some configurations, non-existent.
This analysis covers the doctrinal background, competing interpretations of the existing legislation, the gap between statute and practice. Cross-border implications for European clients. Additionally, the strategic choices available to businesses seeking to protect AI-generated outputs in Ireland today.
Doctrinal foundations: what Irish copyright law says about machine-made works
Irish intellectual property legislation governing copyright was substantially shaped in the late twentieth century. At that time, the concept of a "computer-generated work" referred to outputs of relatively simple algorithmic processes – not to large language models or generative image systems. The statutory provision that addresses such works assigns authorship to the person who makes the necessary arrangements for the work's creation. This provision was prescient in some respects. In others, it is a source of serious doctrinal ambiguity when applied to contemporary AI.
The core difficulty is the word "arrangements." Under traditional software use, a human programmer or operator could plausibly be said to have made all material creative decisions. With modern generative AI, the situation is more complex. The model's internal parameters, trained on vast corpora, produce outputs that no single human directed in any meaningful creative sense. The prompt author, the model developer, the dataset curator, and the deploying business each contribute some element. None contributes what copyright law traditionally calls authorship.
Under Irish copyright doctrine, a work must be original to attract protection. Originality in the Irish context. shaped by EU harmonisation and the Infopaq standard derived from the Court of Justice of the European Union – now requires that a work reflect the author's own intellectual creation. This requirement creates a structural problem. If no human author exercised creative choices in producing the work, the EU-derived originality standard may not be satisfied. The computer-generated works provision, which predates EU harmonisation of this standard, sits in uneasy tension with it. Practitioners in Ireland note that this tension has not been authoritatively resolved.
Irish courts follow a common law tradition of incremental development through case-by-case adjudication. Absent direct authority on AI-generated works, a court would likely reason by analogy from existing decisions on photographs, databases, and software. Each analogy carries different implications. A photograph attracts protection if the photographer exercised creative choices in framing, lighting, or timing. A database may attract a sui generis right based on investment rather than creativity. Software is protected as a literary work. None of these categories maps cleanly onto a generative AI output.
For businesses seeking to build IP portfolios around AI-generated content, this doctrinal instability is not merely academic. It affects licensing, enforcement, insurance, and valuation. A right that is legally uncertain cannot be licensed with confidence, enforced against an infringer without risk of challenge, or accurately valued in a due diligence exercise. The practical consequences of getting this wrong compound over time as AI-generated assets accumulate in a business's portfolio.
For a comprehensive overview of the legal services available to technology companies operating in Ireland, see our AI and technology law practice in Ireland, which covers both regulatory compliance and IP strategy.
Competing interpretations and the gap between statute and practice
Three distinct interpretive positions have emerged among legal specialists addressing AI-generated works in common law jurisdictions including Ireland. Each leads to different practical outcomes.
The first position holds that the computer-generated works provision captures AI outputs and vests rights in the deploying business or the person who configured the system. Under this reading, the statute is technologically neutral: it was designed to handle cases where no human author in the traditional sense was involved, and generative AI is simply the latest instance of that case. The breadth of "arrangements" is read generously to include selecting the AI model, designing the prompt, curating training data, and deploying the output commercially.
The second position holds that the EU-derived originality standard takes precedence and effectively overrides the computer-generated works provision for works produced within the scope of EU copyright directives. Under this interpretation, if no human intellectual creation is expressed in the work, no copyright subsists – regardless of how the domestic provision is drafted. This position has the most support in recent EU-level commentary and aligns with the direction of travel in the Court of Justice's case law on originality. Its practical consequence is that a significant portion of AI-generated output receives no copyright protection at all.
The third position – perhaps the most pragmatic but also the least doctrinally settled – focuses on the degree of human involvement at each stage of the process. Where a human author exercised meaningful creative choices in the prompt, in selecting among candidate outputs, or in editing and arranging the final work, protection may attach to those human-contributed elements. The AI-generated portions remain unprotected; the human-curated whole may qualify. This approach is consistent with how courts in other EU jurisdictions have begun to address the question, and it provides a partial path to protection without requiring legislative change.
In practice, the gap between these interpretations creates a compliance and strategy vacuum. Businesses cannot simply read the statute and know where they stand. The answer depends on the specific AI system used, the degree of human involvement, the nature of the output. And. critically. whether and how an Irish court would apply the EU harmonisation framework to the domestic provision. None of these variables has been litigated in Ireland at the time of writing.
This uncertainty has direct consequences for technology licensing. A company that attempts to license AI-generated outputs to a third party must represent that it holds valid rights in those outputs. If that representation is incorrect – because the works are in the public domain or because the copyright vests elsewhere – the licence is worthless and potential liability for misrepresentation arises. Similarly, a business that seeks to enforce against a competitor copying its AI-generated assets faces the threshold question of whether it has standing to bring the claim at all.
Algorithmic accountability is a related but distinct concern. Even where rights do subsist, a business relying on AI-generated content must be able to demonstrate provenance and authorship in any enforcement action. Modern AI systems do not produce audit trails that map neatly onto copyright ownership chains. The absence of such documentation weakens any claim before an Irish court. Practitioners in Ireland increasingly recommend that clients build contemporaneous records of the human decisions made during the AI generation process. not as a formality. However. As the only available means of establishing the "arrangements" required by the statute.
Software liability intersects with these questions where AI-generated code is deployed in commercial products. Under Irish law, software is protected as a literary work. Code produced by an AI system raises the same originality questions as other AI outputs. If the code is found to be in the public domain, a business that has licensed it exclusively to customers cannot enforce that exclusivity. Worse, if the AI training data included proprietary code without authorisation, the output may carry embedded infringement risks that the deploying business inherits. This risk is material, not theoretical, and it is one that software liability analysis must address before any significant deployment.
Cross-border implications for European clients
Ireland occupies a distinctive position in the European AI and technology regulatory environment. As a common law jurisdiction within the EU, it must implement EU directives while applying domestic law through case law rather than codified civilian doctrine. This creates a dual layer of complexity for businesses operating across European markets.
The EU AI Act, which entered into force in 2024 and is being phased in through to 2027, does not directly address copyright ownership of AI-generated works. It focuses on risk classification, transparency obligations, and systemic risk governance for AI systems. However, it intersects with IP questions in several indirect ways. High-risk AI systems deployed in the EU require documentation of training data and human oversight mechanisms. That documentation, if properly maintained, also serves as evidence of the "arrangements" required for copyright vesting under Irish law. AI Act compliance thus has IP-protective side effects for businesses that implement it rigorously.
The EU's approach to copyright and AI has been addressed in a series of recitals and guidance documents accompanying recent digital legislation. The general direction is toward a requirement of human creative input for copyright protection, combined with a text and data mining exception that permits AI training on lawfully accessed content subject to opt-out regimes. For businesses that trained proprietary AI models in Ireland or deployed models trained elsewhere, the interaction between these regimes and Irish copyright law is not yet settled.
Digital services regulation at EU level adds further complexity. Platforms that host AI-generated content may face different liability exposure depending on whether the content is treated as copyrighted material or as public domain. The liability framework applicable to digital service providers under EU law creates obligations around notice-and-takedown and content moderation that interact with IP status. A platform that incorrectly classifies AI-generated content as unprotected may remove material that is in fact the client's IP asset. One that incorrectly classifies it as protected may face claims from third parties whose training data was used without authorisation.
For businesses operating between Ireland and other EU jurisdictions, the lack of uniform treatment of AI-generated works across member states creates arbitrage risk and strategic opportunity. A work that may not attract copyright in Ireland under the EU originality standard might attract protection in a jurisdiction that applies a lower "skill and labour" standard. Conversely, enforcement of Irish-based rights in a continental EU court may require demonstrating compliance with EU harmonisation requirements that an Irish domestic analysis might overlook. Cross-border IP strategy must therefore account for the most restrictive applicable standard, not merely the domestic one.
The interaction with Irish and EU competition law is also relevant. Where a business claims exclusive rights in a large volume of AI-generated content. for example, a dataset or a body of training material – competition authorities may scrutinise whether those rights create market foreclosure effects. This concern is emerging, not yet well-developed, but it is visible in EU-level regulatory discussions around data access and AI infrastructure. Businesses building significant IP positions in AI-generated content in Ireland should monitor this regulatory trajectory.
For clients who also hold or seek intellectual property rights in other instruments. trademarks, patents. Alternatively. Design rights. see our analysis of the broader intellectual property practice in Ireland. This addresses the interaction between copyright and other IP rights for technology businesses.
To explore how cross-border AI and IP strategy applies to your business in Ireland, contact us at info@ferrazwhitmore.com.
Strategic recommendations for businesses building AI-generated IP in Ireland
The current state of Irish law does not permit businesses to simply rely on the computer-generated works provision and assume their AI outputs are protected. A more structured approach is required. The following recommendations reflect the practical measures available to businesses under existing law, pending statutory clarification or judicial authority.
The first and most important step is documentation. Businesses should maintain contemporaneous records of every human decision made in the AI generation process: the selection of the model. The design of the prompt, the criteria applied in choosing among candidate outputs. Additionally, any editing or curation of the final work. These records serve two functions. They establish the "arrangements" required for the computer-generated works provision. They also demonstrate human intellectual creation sufficient to satisfy the EU originality standard where the human contribution is meaningful. Documentation should be systematic, not ad hoc, and should be built into the production workflow rather than reconstructed after a dispute arises.
The second step is contractual layering. Where statute is uncertain, contract can fill gaps. Businesses should ensure that their agreements with AI service providers, employees, and contractors include clear assignment and work-for-hire provisions covering AI-generated outputs. These provisions should address both the AI outputs themselves and any human-curated selections from those outputs. Technology licensing agreements should include representations and warranties about IP ownership that are qualified to reflect the actual state of the law – not representations that overstate the strength of the rights held.
The third step is risk stratification. Not all AI-generated assets carry the same IP risk profile. A marketing image generated from a stock AI tool carries different risks from a proprietary software module generated by a bespoke model trained on the business's own data. The former may be commercially replaceable if IP protection fails; the latter may represent core technical infrastructure. Businesses should conduct an asset-by-asset review, prioritising human involvement in the production of their highest-value AI-generated outputs and reserving traditional human authorship for works where IP protection is commercially critical.
The fourth step is monitoring the regulatory trajectory. The Irish Government has signalled awareness of the gap between existing copyright legislation and AI realities. At EU level, ongoing consultations on AI and copyright are likely to produce legislative developments within the medium term. Businesses should track these developments and be prepared to adapt their IP strategies as the law crystallises. The window for structuring IP positions under current uncertainty is limited. Once legislative or judicial clarity arrives, the options available will narrow.
The fifth consideration is AI Act compliance as an IP enablement tool. Businesses subject to the EU AI Act's documentation and transparency obligations should treat those obligations as an opportunity, not merely a cost. Robust documentation of AI system design, training data provenance, and human oversight – required for AI Act compliance – also creates the evidential record that best supports an IP ownership claim under Irish law. These two regulatory regimes, though separate in purpose, are complementary in practice.
A non-obvious risk deserves specific attention: the provenance of training data. If an AI system was trained on third-party copyrighted material without authorisation – or without reliance on a valid text and data mining exception – the AI-generated outputs may carry embedded infringement liability. This risk is not cured by the computer-generated works provision, which addresses ownership rather than infringement. A business that deploys such a system and commercialises the outputs may face claims from the original rights holders. Due diligence on AI system provenance is therefore essential before any significant commercial deployment. Many businesses underestimate this exposure precisely because it arises not from their own actions but from decisions made by AI developers upstream.
For a comparative analysis of how these questions are addressed in other civil law jurisdictions, the discussion of AI-generated works and intellectual property in Portugal offers a useful contrast with the Irish common law approach.
To discuss how AI Act compliance and IP documentation strategy apply to your specific situation in Ireland, reach out to info@ferrazwhitmore.com.
Outlook: where Irish law on AI-generated works is heading
The trajectory of Irish law on AI-generated works is unlikely to remain static. Several forces are converging to push toward legislative or judicial clarification within the next few years.
First, the volume of commercially significant AI-generated content in Ireland is growing rapidly. As that volume grows, so does the number of disputes in which IP status becomes outcome-determinative. Irish courts operate within a common law tradition that develops doctrine through decided cases. The first significant Irish court decision addressing AI-generated copyright will establish a precedent that shapes the landscape for years. Businesses with strong positions in AI-generated content have an interest in being ready to litigate those positions effectively when the opportunity arises.
Second, the EU legislative cycle is moving. The AI Act's implementation will generate ancillary guidance and secondary legislation that touches on IP questions. More directly, the European Commission's work on copyright in the digital environment is ongoing. Harmonisation measures that address AI-generated works are a plausible outcome within the medium term. When they arrive, Ireland will be required to implement them. The direction of EU-level thinking – toward human intellectual creation as a prerequisite for copyright – suggests that the computer-generated works provision's current breadth may narrow rather than expand under harmonisation.
Third, international pressure is building. Major trading partners – including the United Kingdom, the United States, and key Asian jurisdictions – are all grappling with the same questions. The UK Intellectual Property Office has been consulting on whether and how to extend protection for AI-generated works. US copyright authorities have taken a position that AI-generated works without human authorship do not attract federal copyright protection. These positions create pressure on Irish law to clarify its own stance, particularly given Ireland's role as a base for major technology companies operating across all three markets.
For businesses, the practical implication is that rights secured under current Irish law may be affected by future legislative change. A position built on the computer-generated works provision today may need to be rebuilt around human authorship requirements tomorrow. This is not a reason to delay structuring IP positions. It is a reason to structure them in a way that maximises human involvement now, creating a fallback position that survives even under a stricter future standard.
The businesses best placed to benefit from eventual clarity are those that have already invested in documentation, contractual protections, and human-centric workflows. Those that relied on broad assumptions about automatic ownership will face the most difficult transition when judicial or legislative authority arrives. The lost opportunity in this area is not simply failing to protect existing assets. It is failing to build the evidentiary and contractual infrastructure that makes those assets protectable at all.
Frequently asked questions
Q: Does Irish law currently protect AI-generated works under copyright?
A: Irish copyright legislation contains a provision for computer-generated works that could, in principle, cover AI outputs. However, EU harmonisation requirements impose an originality standard based on human intellectual creation that may not be satisfied where no human author made meaningful creative choices. The answer is legally uncertain and untested in Irish courts. Businesses should not assume automatic protection and should take active steps to document human involvement in their AI generation workflows.
Q: How long does it take to put in place a practical IP protection strategy for AI-generated content in Ireland?
A: A focused review of existing AI-generated assets, combined with the introduction of documentation protocols and contractual protections, can typically be completed within several weeks for a business of moderate complexity. More extensive due diligence on AI system provenance – particularly training data – may take longer. The sooner a business begins, the stronger the evidentiary record it builds. Waiting for legislative clarity before acting means accumulating a backlog of unprotected or poorly documented assets.
Q: A common misconception is that using an AI tool automatically means the business owns all outputs – is this correct?
A: This is one of the most frequent and costly misconceptions in AI-related IP matters. Engaging a lawyer in Ireland with specific experience in technology licensing and copyright will reveal that ownership depends on the degree of human creative involvement. The terms of the AI provider's own licence, the training data provenance. Additionally, the interaction between domestic and EU law. Simply paying for access to an AI service does not vest copyright in the outputs. The AI provider's terms of service may themselves limit, exclude, or qualify any rights the user might otherwise claim.
About Ferraz & Whitmore
Ferraz & Whitmore is an international law firm based in Lisbon, advising business clients across 46 jurisdictions. Our practice combines Portuguese civil law expertise with English common law tradition. a dual foundation that is particularly well-suited to advising on AI and technology law matters in Ireland. There. EU regulatory obligations must be addressed through a common law institutional context. As a law firm in Ireland with cross-border capabilities, we advise technology companies, institutional investors. Additionally, in-house legal teams on AI Act compliance. Algorithmic accountability, software liability, technology licensing. Additionally, the protection of AI-generated intellectual property. Our intellectual property practice covers digital services, copyright, and cross-border enforcement across European and international markets. The firm's attorneys have advised on technology transactions and IP matters across both civil law and common law systems, with direct experience before EU-level regulatory processes. To discuss how Irish and EU law applies to your AI-generated IP, contact us at info@ferrazwhitmore.com.
Disclaimer: This publication is provided for informational purposes only and does not constitute legal advice. The information herein should not be relied upon as a substitute for professional legal counsel tailored to your specific circumstances. Ferraz & Whitmore assumes no liability for actions taken or not taken based on the contents of this material. For advice regarding your particular situation, please contact info@ferrazwhitmore.com.