A European technology company deploys a generative AI system to produce marketing content, product images, and software code. The output is commercially valuable. The question of who owns it – and whether it can be protected at all – is not rhetorical. In Greece, as across much of the European Union, the answer is genuinely unsettled. Greek intellectual property legislation was written when the author was assumed to be a human being. It was not designed for outputs produced autonomously by machine learning systems. Businesses that treat AI-generated content as freely protectable risk discovering, too late, that a competitor has successfully contested their ownership claim.
AI-generated works in Greece occupy a doctrinal gap: existing intellectual property legislation ties copyright to human authorship and original creative expression, leaving outputs produced without direct human creative input in a legally ambiguous position. Businesses can reduce this exposure by structuring human involvement carefully and by documenting the contribution of natural persons at each stage of the AI-assisted production process. The emergence of EU AI Act compliance obligations adds a parallel regulatory dimension that companies operating in Greece must address alongside – but separately from – intellectual property strategy.
This analysis covers the doctrinal foundations of Greek copyright doctrine, competing interpretations emerging in court and regulatory practice. The practical gap between statutory text and day-to-day commercial reality, cross-border implications for European clients. Additionally, the strategic steps that businesses can take now to protect their position.
Doctrinal foundations: human authorship in Greek intellectual property law
Greek intellectual property legislation is rooted in the civil law tradition. It establishes copyright as a right that attaches to the natural person who creates a work through an act of intellectual effort. The concept of originality in Greek law is closely tied to the author's own creative choices – the selection, arrangement, and expression that reflect individual intellectual personality.
This approach mirrors the broader European position articulated through EU harmonisation directives covering software, databases, and literary and artistic works. Greek courts, in line with the Areios Pagos (Supreme Civil and Criminal Court of Greece), have consistently interpreted originality as requiring a creative act by a human being. A work qualifies for protection when it bears the intellectual imprint of its author. A work produced entirely by a machine – without meaningful human creative choice at the point of generation – does not, under this reading, satisfy the threshold.
The consequence for AI-generated content is direct. If a natural person does nothing more than submit a prompt to a generative AI system and receive output, the resulting text, image, or audio track may not attract copyright protection under Greek law. The human contribution – selecting a tool, defining parameters, pressing a button – may be insufficient to constitute the creative act that Greek intellectual property legislation contemplates.
This is not a hypothetical concern. It is the position that a well-advised competitor, or a party seeking to copy AI-generated content, would advance in any Greek court proceeding.
Competing interpretations and the originality threshold
The doctrinal picture is not entirely settled, however. Practitioners in Greece note that courts have not yet produced a clear, definitive line on AI-assisted works. Two interpretive currents can be identified.
The first – and currently dominant – reading applies the traditional human-authorship requirement strictly. Under this approach, AI-generated output receives no copyright protection unless a human author made creative choices that are genuinely reflected in the final work. The AI system is treated as a sophisticated tool, analogous to a camera or a word processor. Ownership, if it exists at all, vests in the person whose creative decisions shaped the output – not in the person who merely operated the system.
The second reading takes a more functional approach. It focuses less on where in the production chain the creative act occurred and more on whether the overall process, taken as a whole, involved sufficient human intellectual engagement to justify protection. A human who designs the training dataset, selects the model architecture, iterates across multiple outputs, and curates the final result may, on this view, have made enough creative choices to qualify as an author. This reading finds some doctrinal support in the treatment of computer-generated works in other EU member states and in the broader trajectory of EU harmonisation in digital services and technology licensing.
Greek courts have not definitively adopted either position. The Areios Pagos has addressed originality in the context of conventional creative works and software, but AI-specific authorship disputes have not yet produced settled case law. This gap creates both risk and opportunity. A business that structures its AI-assisted production process to maximise identifiable human creative contribution is better positioned to assert protection. A business that simply harvests raw AI output and relies on ownership assumptions may find that protection is unavailable.
The distinction between human-assisted AI output and fully autonomous AI output is therefore not merely academic. It is the operative variable that determines whether a body of commercially valuable content is protectable at all.
To explore how AI law and intellectual property intersect in the Greek regulatory environment, see our dedicated practice page on AI and technology law in Greece.
The gap between statute and practice: what actually happens in disputes
Even where the doctrinal position is reasonably clear, a significant gap separates the text of Greek intellectual property legislation from the practical reality of disputes involving AI-generated content.
The first practical difficulty is evidentiary. Greek civil procedure requires a claimant asserting copyright to establish authorship. For AI-assisted works, this means demonstrating that a natural person made creative choices reflected in the work. Many businesses cannot do this because they have not documented the human involvement in their AI-assisted production processes. They cannot identify, after the fact, which prompt iterations, curation decisions, or editorial interventions shaped the final output. The absence of documentation is not a minor inconvenience – it is frequently fatal to an infringement claim.
The second difficulty involves software liability and the question of who bears responsibility when an AI system produces output that infringes a third party's rights. Greek civil liability rules, read alongside EU digital services legislation, do not produce a clean answer. The developer of the AI model, the business deploying it, and the end user each occupy different positions in the liability chain. Greek courts applying general civil liability principles will examine who had control over the system and who had the capacity to prevent the infringing output. In practice, this analysis is highly fact-specific and frequently contested.
Technology licensing agreements add a further layer of complexity. Many AI systems are deployed under commercial licences that purport to assign ownership of generated outputs to the user, or that disclaim any warranty of non-infringement. Whether these contractual provisions are enforceable against third-party copyright holders. whose rights are not subject to private agreement between licensor and licensee – is a separate question that Greek courts have not yet systematically addressed.
A non-obvious risk arises in the training data dimension. If an AI system was trained on copyrighted material without a licence, the outputs it produces may themselves be tainted by the underlying infringement. A Greek business that purchases or deploys an AI system without conducting due diligence on the training data provenance may face infringement exposure that it did not originate and did not anticipate. Practitioners in Greece note that this risk is particularly acute for businesses in creative industries – music, publishing, visual arts, software development – where the probability that training data included protected material is high.
For a detailed treatment of intellectual property protection strategies available to businesses operating in Greece, see our analysis of intellectual property law in Greece.
To explore how comparable questions have been addressed in the Portuguese legal system – a civil law jurisdiction facing analogous doctrinal tensions – see our deep analysis of AI-generated works and intellectual property in Portugal.
To receive a tailored assessment of your AI-generated content ownership position in Greece, contact us at info@ferrazwhitmore.com.
Cross-border implications and the EU AI Act dimension
For European businesses operating across multiple jurisdictions, the Greek position on AI-generated works cannot be considered in isolation. It sits within an evolving EU legislative regime that is reshaping the conditions under which AI systems are deployed – and, indirectly, the conditions under which AI-generated content is produced and exploited.
AI Act compliance obligations are now a live commercial and legal reality. The EU AI Act classifies AI systems by risk level and imposes obligations on both providers and deployers. For general-purpose AI models – the large language and image generation systems most commonly used to produce AI-generated content – the Act introduces transparency requirements and algorithmic accountability obligations. Providers of high-capability general-purpose models must maintain technical documentation, implement copyright policies, and publish summaries of training data used.
These obligations do not directly resolve the question of who owns AI-generated output. They operate on a different legal track. But they have significant indirect relevance. A business that can demonstrate, through AI Act compliance documentation. That its AI system was deployed in accordance with regulatory requirements. and that its training data policies were lawful. is better positioned to defend against training-data-based infringement claims. Algorithmic accountability records may also assist in establishing the extent of human creative involvement in the production process.
The interaction between Greek intellectual property law and EU harmonisation directives is also evolving. The EU's approach to copyright in the digital environment. particularly through legislation on digital single market rules – has introduced text and data mining exceptions that affect how AI systems can lawfully process copyrighted material. Greek domestic law has implemented these EU rules. A business whose AI system relies on text and data mining of copyrighted content must verify that its activities fall within the permitted scope of those exceptions. Additionally. That the rights holders whose content was used had not lawfully opted out.
Cross-border enforcement adds further complexity. A work produced by a business in Greece but distributed across the EU may be subject to copyright claims in multiple member states simultaneously. The applicable law rules of EU private international law determine which national copyright regime governs each aspect of the dispute. Businesses that assume Greek law exclusively governs their AI-generated content may face parallel proceedings under German, French. Alternatively. Dutch intellectual property legislation. each of which may apply a different originality threshold or a different approach to AI-assisted authorship.
The civil law tradition that Greece shares with Portugal, France, Spain, and other continental European jurisdictions creates a degree of doctrinal consistency. But the gap between those jurisdictions' treatment of AI authorship is not negligible. A business managing a pan-European content portfolio needs jurisdiction-specific advice, not a single EU-wide assumption.
For a preliminary review of your cross-border AI content strategy across Greek and EU jurisdictions, email info@ferrazwhitmore.com.
Strategic recommendations and self-assessment checklist
Given the doctrinal uncertainty, businesses generating or exploiting AI-assisted content in Greece should take a structured approach to ownership, protection, and risk management. The following considerations define the practical strategy.
Maximise and document human creative contribution. The single most effective step a business can take is to ensure that human authors make identifiable, documented creative decisions at multiple stages of the AI-assisted production process. This means recording prompt design, curation choices, iterative editing, and final selection decisions. The documentation does not need to be elaborate – it needs to be sufficient to demonstrate, in litigation, that a natural person's creative choices are reflected in the final output.
Audit AI system licences and training data provenance. Before deploying any AI system to produce commercially valuable content, conduct due diligence on the licence terms and on the provenance of the training data. Verify whether the licence purports to assign output ownership to the user and whether that assignment is enforceable in Greece. Identify whether the system's training data was obtained under lawful text and data mining conditions or under a licence from rights holders.
Use notarial and deposit mechanisms to establish date of creation. Because copyright in Greece arises automatically and without registration, disputes frequently turn on the date of creation and the identity of the author. Businesses can strengthen their evidentiary position by using notarial deeds or third-party deposit services to create dated records of AI-assisted outputs and the human choices that shaped them.
Address AI Act compliance as a parallel obligation. Map your AI system deployments against the risk classifications and transparency requirements of the EU AI Act. Maintain algorithmic accountability records. Implement copyright policies for general-purpose AI model use. Treat compliance documentation as a dual-purpose asset – it satisfies regulatory obligations and may support intellectual property claims.
This approach is applicable if:
- Your business produces commercially valuable content using generative AI systems.
- Your content is distributed in Greece or other EU member states.
- Your AI system relies on training data that may include third-party copyrighted material.
- Your contractual arrangements with AI providers have not been reviewed against Greek intellectual property and civil liability rules.
Before proceeding, verify:
- Whether human creative choices in your AI-assisted production process are documented.
- Whether your AI system licence assigns output ownership in terms enforceable in Greece.
- Whether your training data use falls within lawful text and data mining exceptions.
- Whether your AI Act compliance obligations have been assessed and addressed.
When any of the above conditions are unmet, the business faces a dual exposure. It may lose the ability to enforce ownership over its AI-generated content. It may simultaneously face infringement claims from third parties whose rights were engaged in the production process. These risks do not disappear with the passage of time. they crystallise when a competitor copies the content, when a rights holder makes a claim. Alternatively. When a business seeks to assign or licence its content portfolio and a buyer's legal due diligence surfaces the unresolved ownership questions.
Outlook: where Greek and European law is heading
The legal treatment of AI-generated works in Greece is unlikely to remain static. Several developments will shape the position over the next several years.
At the EU level, the AI Act's implementation timeline will bring greater regulatory clarity on the obligations of AI system providers and deployers. As compliance infrastructure matures, the documentation and accountability records that businesses generate for regulatory purposes will increasingly serve as evidence in intellectual property disputes. This convergence between AI regulation and intellectual property practice is one of the defining features of the emerging legal environment.
At the national level, Greek courts will eventually produce decisions on AI-assisted authorship. The first cases are likely to involve high-value commercial disputes – software development, visual content, music generation – where the stakes justify the cost of litigation. Those decisions will anchor the doctrinal position and reduce the current uncertainty. Businesses that have structured their processes correctly will be well placed when that case law emerges. Businesses that have not may find themselves on the wrong side of the first significant precedent.
The broader European debate on whether to introduce a specific copyright category for AI-generated works – analogous to the approach taken in the United Kingdom for computer-generated works – remains open. Several EU member states have advocated for harmonised treatment. The European Commission has signalled awareness of the gap. Whether harmonisation will be achieved through targeted copyright legislation, through guidance from the Court of Justice of the European Union, or through the accumulated weight of national case law is not yet clear.
What is clear is that the window for proactive structuring is open now. Businesses that invest in documentation, licence due diligence, and AI Act compliance today will have a stronger position when the law settles. Those that wait for certainty before acting may find that the settled position is less favourable than the steps they could have taken when the question was still open.
Frequently asked questions
Q: Can an AI system be recognised as the author of a creative work under Greek intellectual property law?
A: No. Greek intellectual property legislation, consistent with the broader European civil law tradition, requires that authorship attach to a natural person whose creative choices are reflected in the work. An AI system lacks legal personality and cannot hold copyright. The operative question for businesses is not whether the AI is an author, but whether the human contribution to an AI-assisted output meets the originality threshold that Greek courts apply.
Q: How long does it typically take to obtain copyright protection for an AI-assisted work in Greece, and what costs are involved?
A: Copyright in Greece arises automatically upon the creation of an original work – no registration is required, and there is no official filing fee for copyright itself. However, businesses that want enforceable evidence of authorship and creation date commonly use notarial deeds or deposit services, which carry notarial and administrative costs in the range of hundreds of euros. Disputes over AI-assisted authorship, if they reach the civil courts, can take one to three years to resolve at first instance.
Q: Does the EU AI Act change IP ownership rules for AI-generated content in Greece?
A: A common misconception is that AI Act compliance automatically resolves intellectual property questions. The two regimes operate on separate legal tracks. The AI Act imposes transparency and risk-management obligations on AI system providers and deployers – including algorithmic accountability requirements – but it does not amend copyright doctrine or alter ownership rules. Businesses must address both dimensions independently: satisfying AI Act compliance obligations does not confer or confirm copyright ownership over AI-generated outputs. Engaging a lawyer in Greece with cross-border experience in both regulatory compliance and intellectual property is the most effective way to address both dimensions simultaneously.
About Ferraz & Whitmore
Ferraz & Whitmore is an international law firm based in Lisbon, advising business clients across 46 jurisdictions. Our team combines Portuguese civil law expertise with English common law tradition to deliver cross-border legal solutions in AI and technology law, intellectual property protection, and digital services regulation. In Greece, we advise technology companies, platform operators, and institutional investors on AI Act compliance, algorithmic accountability, software liability, and technology licensing arrangements for AI-assisted content. Our intellectual property practice covers copyright, software protection, and trade mark strategy across EU and non-EU jurisdictions. As a law firm in Greece with a Lisbon base, we provide direct access to EU regulatory developments while supporting enforcement and arbitration strategies in common law jurisdictions. Our attorneys have advised on AI and intellectual property matters across both civil law and common law systems, including in jurisdictions where the treatment of AI-generated works diverges significantly from the Greek position. The firm's AI law practice participates in cross-border practice groups focused on technology regulation and digital services. To discuss your AI-generated content strategy in Greece, 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.