A media company based in Warsaw commissions a series of marketing visuals using a generative AI platform. The outputs are striking, commercially valuable, and produced at a fraction of traditional cost. Then a competitor reproduces them without authorisation. The company's legal team discovers, at the worst possible moment, that it may hold no enforceable copyright in any of those images. This scenario is no longer hypothetical. It is the central IP challenge confronting technology-driven businesses operating in Poland today.
Under Polish intellectual property legislation, copyright protection attaches to works that reflect the personal creative expression of a natural person. AI-generated outputs currently sit outside this protection unless a qualifying degree of human creative contribution can be demonstrated. Businesses that fail to document that contribution risk losing enforceable rights in content they have invested significant resources to produce.
This analysis examines the doctrinal position under Polish law, competing interpretations emerging in legal practice, the gap between statute and actual enforcement. Cross-border implications under EU law. Additionally, the strategic steps that international businesses can take to protect their interests now.
Doctrinal foundations: what Polish intellectual property law says about authorship
Polish intellectual property legislation has remained formally unchanged since its consolidation in the 1990s. Its core authorship doctrine is unambiguous: a work is protected if it constitutes a manifestation of creative activity of an individual character and is the product of a natural person. This formulation has served Polish courts well for decades. It was drafted with human authors in mind and makes no provision for machine-generated content.
The concept of individual character (indywidualny charakter – the requirement of personal creative imprint in Polish copyright doctrine) is the decisive threshold. Courts in Poland assess whether a work bears the author's personal intellectual stamp. A generative AI system, however sophisticated, cannot hold subjective experience or personal perspective. The output of such a system therefore fails the personal-imprint test on its face.
This does not mean that all AI-assisted content is unprotectable. Polish intellectual property legislation draws an implicit but legally significant distinction between works created with AI tools and works created by AI systems. The former can attract protection when a human author makes genuine creative choices – selecting prompts, curating outputs, arranging elements, applying editorial judgment. The latter, where the system operates autonomously without meaningful human direction, produces outputs that are likely unprotected.
The challenge is that the line between these two categories is blurry. No Polish court has yet issued definitive guidance on precisely how much human input is required. Practitioners in Poland note that this uncertainty is not unique to the jurisdiction. it mirrors debates across all EU member states. but it carries particular commercial risk in a market where AI content tools have been adopted rapidly across the media, advertising, and software sectors.
Software itself occupies a distinct doctrinal position. Under Polish intellectual property legislation, computer programs are protected as literary works. The source code of an AI system may therefore attract copyright held by its developer. This is well-settled law. What remains unsettled is whether the output of that program – the image, the text, the musical composition – benefits from any derivative protection. Current doctrine says no, unless a human author's contribution can be independently identified in the output itself.
Competing interpretations and the gap between statute and practice
Legal practice in Poland has developed several competing approaches to the authorship question, none of which has yet achieved doctrinal dominance. Understanding each is essential for any business structuring its AI content strategy.
The first approach treats AI outputs as unprotected res nullius – things belonging to no one. Under this view, a fully AI-generated image or text enters the public domain immediately upon creation. Any person may use, reproduce, or adapt it without restriction. This approach is legally coherent and reflects the plain text of Polish intellectual property legislation. Its commercial consequence is severe: competitors can freely copy AI-generated marketing materials, product descriptions, or design assets.
The second approach focuses on the prompt engineer or operator. Proponents argue that the human who designs, refines, and iterates a prompt exercises sufficient creative direction over the AI to qualify as an author of the resulting output. This interpretation is contested. Polish intellectual property doctrine has never extended authorship to persons who merely instruct a tool, as opposed to those who directly produce expressive content. A photographer who directs a subject is an author; the subject is not. Whether the analogy extends to AI prompting remains an open question for Polish courts.
The third approach, increasingly favoured by practitioners advising technology companies, focuses on the post-generation selection and arrangement of AI outputs. If a human editor chooses among multiple AI-generated variants, combines elements, applies corrections, and makes substantive editorial decisions, that editor's contribution may be sufficient to establish authorship in a curated compilation or derivative work. This is the safest practical position under current Polish law, even though it requires documented human involvement at every stage.
The gap between statute and practice also manifests in technology licensing. When businesses acquire licences to use generative AI platforms, those licences typically address the provider's liability but are often silent on the IP status of outputs. A technology licensing agreement that does not explicitly allocate output rights leaves both parties in a legally ambiguous position. Polish civil procedure rules permit courts to interpret contractual silence by reference to the reasonable expectations of the parties and the nature of the transaction. but that interpretive process introduces uncertainty and litigation risk that could be avoided by clear contractual drafting.
Algorithmic accountability is a related dimension. Where an AI system generates content that infringes a third party's pre-existing copyright – by reproducing training data too closely – the question of software liability becomes acute. Polish courts have not yet resolved whether liability rests with the AI developer, the operator who deployed the system, or the user who prompted the generation. The answer will likely depend on the degree of control each party exercised over the system and the foreseeability of the infringing output. Specialists in technology law note that the absence of clear software liability rules in this context is one of the most commercially significant gaps in Poland's current digital services regulatory environment.
For a tailored strategy on AI-generated content rights and technology licensing in Poland, reach out to our AI and technology law practice in Poland to discuss your specific situation.
The EU AI Act dimension: compliance as an IP strategy tool
The EU AI Act entered into force in stages beginning in 2024 and applies directly in Poland as EU regulation. Its obligations are primarily framed around risk classification, transparency, and documentation of AI systems. However, AI Act compliance has an underappreciated secondary effect: it creates an evidential record that businesses can deploy in IP disputes.
General-purpose AI models – the large language models and image-generation systems most commonly used to produce commercial content – are subject to specific transparency and documentation requirements under the AI Act. Providers of such models must maintain technical documentation, implement copyright policies, and publish summaries of training data. For Polish businesses that use these models, this documentation becomes a resource. It can help establish which elements of an AI output derive from the provider's training choices and which reflect the user's creative direction.
More directly, a business that builds AI Act compliance into its content production workflow. documenting human creative decisions at each stage, maintaining logs of editorial choices. Retaining version histories. simultaneously builds the evidentiary foundation for an authorship claim under Polish intellectual property legislation. Compliance and IP protection become the same process, executed once.
The AI Act also introduces obligations relevant to deepfakes (AI-generated synthetic media) and to AI systems used in generating content for the public. Where a Polish business deploys such systems in its digital services, it must implement labelling and transparency measures. Failure to do so exposes the business to regulatory enforcement. and may also undermine its credibility in any subsequent IP dispute. Since non-compliant deployment suggests a lack of the careful human oversight that authorship claims require.
Practitioners in Poland and across the EU increasingly advise clients to treat AI Act compliance not as a cost centre but as an investment in IP infrastructure. The documentation required for regulatory purposes is the same documentation that supports rights claims, technology licensing negotiations, and due diligence processes in M&A transactions involving AI-generated content assets.
For a comparative perspective on how other EU jurisdictions handle the same questions, the analysis of AI-generated works and intellectual property in Portugal identifies both convergences and divergences that are relevant to cross-border portfolio strategy.
Cross-border implications for European businesses operating in Poland
Poland is a significant market for technology businesses expanding eastward within the EU. It is also a jurisdiction where the gap between the pace of AI adoption and the pace of legal adaptation is particularly visible. For European businesses operating cross-border, this creates specific risks that differ from those encountered in Western European markets.
First, the EU information society legislation harmonises certain aspects of copyright across member states, including the exceptions relevant to text and data mining used in AI training. However, it does not harmonise the authorship threshold. Each member state determines for itself what level of human creative contribution qualifies for protection. A work that attracts copyright in one EU jurisdiction may not attract it in another. A business that treats its AI-generated content portfolio as uniformly protected across the EU is operating on a false assumption.
Second, enforcement diverges significantly. Polish courts handling IP matters sit within a civil law system that gives substantial weight to codified doctrine and relatively less weight to judicial creativity in gap-filling. A practitioner accustomed to the English common law tradition – where courts have historically been more willing to develop new IP categories through case law – will find the Polish approach more restrictive in practice. This means that commercial strategies which might work in UK-based licensing arrangements carry higher legal risk when applied to Polish-law contracts.
Third, the cross-border exploitation of AI-generated content raises questions about which law governs. If a Polish business creates AI-generated content using a US-based AI platform, sells licences to a German client. Additionally. The content is ultimately distributed in France, the applicable IP law may vary depending on which aspect of the dispute is in question. The Rome II Regulation provides general guidance on applicable law for IP infringement in cross-border EU disputes. However. Its interaction with AI-specific questions has not yet been tested by the Trybunał Sprawiedliwości Unii Europejskiej (Court of Justice of the European Union).
Fourth, technology licensing agreements involving AI-generated content often cross multiple jurisdictions. A licence granted by a Polish entity to a European counterpart must address the possibility that the licensed content is unprotected in the licensee's jurisdiction. This requires explicit contractual allocation of risk, representations about the degree of human creative involvement, and indemnity provisions that reflect the genuine legal uncertainty.
Specialists in cross-border IP note that the businesses best positioned to manage these risks are those that have built jurisdiction-specific IP audits into their content production processes. An audit conducted by a law firm in Poland with cross-border expertise can identify which elements of an AI-assisted content portfolio meet Polish authorship standards and which require additional human creative input before they can be registered, licenced, or enforced.
For a broader view of intellectual property protection strategies available to international clients operating in Poland, the intellectual property law services page for Poland outlines the procedural options across copyright, trade mark, and related rights.
To explore legal options for protecting AI-generated content assets across EU jurisdictions, schedule a consultation at info@ferrazwhitmore.com.
Strategic recommendations and self-assessment checklist
The lost-opportunity dimension of this area of law is often underestimated. Businesses that fail to structure their AI content production correctly do not merely face uncertain protection – they actively forgo rights that, with modest process adjustments, they could hold. The following strategic positions reflect current best practice under Polish intellectual property legislation and the emerging AI Act compliance environment.
Document human creative contribution at every stage. This is the single most important step. Maintain version histories, prompt logs, editorial decision records, and selection rationales. These documents are not administrative overhead. They are the evidentiary foundation for any future authorship claim or technology licensing negotiation.
Audit existing AI content portfolios before asserting rights. Many businesses have been producing AI-assisted content for months or years without legal review. An audit conducted by a lawyer in Poland with IP expertise can identify which assets are likely protectable, which are at risk, and what remedial steps are available. Starting enforcement action without this audit is a significant strategic error.
Draft technology licensing agreements with explicit output provisions. Any agreement governing the use of AI tools should specify who owns the outputs. What representations the tool provider makes about training data compliance. Additionally, how liability is allocated if the outputs infringe a third party's rights. Silence on these points is not neutral – it creates ambiguity that courts will resolve against the party that drafted the agreement.
Integrate AI Act compliance into content production workflows now. The AI Act's documentation and transparency obligations will tighten over the coming years. Businesses that build compliant workflows today avoid both regulatory exposure and the cost of retrospective remediation. The compliance infrastructure also supports IP rights claims, making it doubly valuable.
Consider trade mark and design rights as supplementary protection. Where copyright is uncertain, trade mark registration of distinctive AI-generated logos. Product identifiers. Alternatively, brand elements may provide enforceable rights that do not depend on authorship doctrine. Polish trade mark legislation and EU trade mark regulation offer parallel registration routes that a law firm in Poland can advise on alongside copyright strategy.
This approach to AI-generated content in Poland is applicable if the following conditions are met:
- The business produces or commissions AI-generated content for commercial use in Poland or for Polish-law licensing arrangements
- The content has commercial value that would be undermined by competitors' free reproduction
- The business has not yet conducted a formal IP audit of its AI content production process
- Existing technology licensing agreements were drafted without specific AI output provisions
- The business operates across multiple EU jurisdictions and needs a consistent IP strategy
Before initiating any enforcement or licensing action. Verify the following critical points: the degree of documented human creative input in each contested work. the governing law provisions in all relevant technology licensing agreements. the AI Act risk classification of the systems used. and whether any third-party training data claims could undermine the business's own rights position.
Frequently asked questions
Q: Can an AI system be named as the author of a work under Polish intellectual property law?
A: No. Polish intellectual property legislation requires that an author be a natural person whose creative activity produced the work. An AI system cannot hold authorship, and a work generated entirely without human creative input currently lacks copyright protection in Poland. The practical consequence is that such works enter the public domain from the moment of creation.
Q: How does the EU AI Act affect IP strategy for AI-generated content in Poland?
A: AI Act compliance requirements – particularly transparency and documentation obligations for general-purpose AI models – create new leverage for IP rights holders. In Poland, a business that documents its human editorial and curatorial decisions over AI outputs builds a stronger claim to copyright protection than one that relies on fully automated generation. Embedding compliance workflows directly into content creation processes is therefore both a regulatory and an IP strategy.
Q: What is the most common mistake international companies make when deploying AI content tools in Poland?
A: Engaging a lawyer in Poland with cross-border IP experience consistently reveals the same error: companies assume that copyright in AI-generated outputs vests in them automatically by virtue of owning or licensing the AI tool. Polish intellectual property legislation does not support this assumption. Ownership of the tool does not confer authorship of its outputs. Companies must document human creative contribution at each production stage if they wish to hold enforceable rights.
About Ferraz & Whitmore
Ferraz & Whitmore is an international law firm based in Lisbon, advising business clients across 46 jurisdictions. Our IP and technology law practice supports international companies operating in Poland on AI-generated content strategy, AI Act compliance, technology licensing, software liability, and cross-border IP enforcement. We combine Polish civil law expertise with English common law tradition to deliver integrated advice for businesses operating across multiple EU legal systems. The firm's IP team includes practitioners with experience before Polish courts and EU-level regulatory bodies, and our Lisbon base provides direct access to EU regulatory developments as they affect digital services across member states. As an international law firm in Poland and across Europe, Ferraz & Whitmore advises technology companies, institutional investors, and in-house legal teams who need results-oriented counsel on emerging technology law questions. To discuss your AI content strategy and IP protection needs in Poland, 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.