A technology company based in Kyiv trains a generative model on a corpus of Ukrainian literary works and produces thousands of images and texts for a commercial platform. The output earns revenue. A rights holder whose original work formed part of the training data files a claim. The platform's investors – several of them foreign – suddenly need to know who owns what, who is liable for what, and whether their technology licensing agreements will hold. Ukrainian intellectual property legislation was not written with this scenario in mind. Neither were the courts.
Under Ukrainian intellectual property legislation, authorship is reserved for natural persons, meaning AI systems cannot hold copyright. Works generated with AI assistance may qualify for protection if a human contributor demonstrates sufficient creative input, but no bright-line statutory threshold exists. Disputes over AI-generated works in Ukraine are resolved under general copyright principles developed for human authors, producing results that are fact-specific and, at present, unpredictable.
This analysis examines the doctrinal foundations of Ukrainian copyright law as they apply to AI output, surveys the gap between the statute and emerging court practice. Addresses cross-border implications for CIS-connected businesses. Additionally, offers strategic recommendations for companies whose commercial models depend on AI-generated content.
Doctrinal foundations: how Ukrainian copyright law approaches authorship and originality
Ukrainian intellectual property legislation inherited the civil law tradition of author-centred copyright. Protection attaches at the moment of creation. It requires no registration, no formality, and no publication. The single condition is originality – conventionally understood as the product of the author's own intellectual effort, reflecting individual creative choices.
This structure creates an immediate tension with AI-generated output. An AI model does not exercise intellectual effort in the legal sense. It applies statistical patterns derived from training data to produce text, images, code, or music. The question is whether a human sufficiently upstream in that process – the developer, the trainer, the prompter, or the operator – can claim the creative contribution that the law demands.
Ukrainian copyright doctrine has not yet produced a definitive answer. Several competing constructions circulate among practitioners. The first treats the AI as a sophisticated tool, analogous to a camera or a word processor. Under this reading, the human who makes the expressive choices – selecting the prompt, curating the output, editing the result – is the author. The second construction holds that where the AI operates autonomously and the human's contribution is merely operational, no protectable authorship exists at all. The output falls into the public domain.
A third, minority view draws on the concept of a tvir, stvorenyi za dopomohoyu shtuchnoho intelektu (work created with the assistance of artificial intelligence) as a distinct category requiring bespoke legislative treatment. This position is intellectually coherent but currently has no statutory basis in Ukraine.
The originality threshold matters enormously in practice. Ukrainian courts examining copyright claims by human authors have generally applied a low originality bar – closer to the continental European minimum than to the higher creative-spark standard found in some common law systems. Whether that low bar applies consistently to AI-assisted works remains unresolved. A client accustomed to English common law's relatively pragmatic approach to computer-generated works will find that Ukrainian civil law provides less statutory certainty, not more.
Software liability adds another layer. The entity that deploys an AI system for commercial content generation may face claims not only under copyright legislation but also under general civil liability rules. Where the AI output reproduces protected expression from training data, the deployer's exposure under software liability doctrine could run in parallel with any copyright infringement analysis.
The gap between statute and practice: what Ukrainian courts have done so far
Ukrainian courts have not yet developed a settled body of doctrine specific to AI-generated works. The cases that have arisen tend to borrow analytical tools from pre-AI copyright disputes and apply them by analogy. The results are instructive, though not yet authoritative.
Courts in Ukraine have generally affirmed that the Verkhovna Rada (Ukrainian parliament) has not enacted AI-specific copyright provisions. Judges therefore work within the existing intellectual property legislation, asking whether a natural person made a qualifying creative contribution. Where plaintiffs can demonstrate that a human author directed the AI in ways that reflect individual aesthetic or expressive choices. selecting among generated options. Arranging outputs into a larger work. Alternatively, substantially editing raw AI output. courts have shown willingness to recognise protection.
Where the human contribution is harder to characterise as creative – for example, entering a brief generic prompt and accepting the first output without modification – the analysis becomes less favourable. Practitioners note that courts have begun scrutinising the degree of human editorial control as a proxy for creative authorship. This is a workable approach in contested litigation. It is, however, deeply fact-sensitive. Two companies using the same generative tool may reach opposite outcomes depending on how their internal production workflows are documented.
A related issue concerns the training data itself. Ukrainian copyright legislation does not provide an explicit text-and-data-mining exception of the kind being introduced across the EU under AI Act compliance frameworks. Rights holders whose works were used in training without authorisation have, in principle, viable claims under existing infringement rules. The practical challenge is evidentiary: establishing that a specific protected work was included in a training dataset. Additionally. That identifiable output reproduces protected expression from that work, requires technical analysis that Ukrainian courts are still developing methodologies to assess.
Algorithmic accountability presents a further dimension. Where an AI system produces output that causes harm – defamatory content, content that infringes third-party rights, or content that misleads consumers – questions of liability attribution arise. Ukrainian civil legislation on general tort principles provides a starting point. But the chain of causation in AI-generated harm is non-standard: developer, trainer, operator, and end user may each bear some share of responsibility. No Ukrainian court has yet articulated a multi-party liability allocation model specific to AI systems.
For businesses operating digital services that rely on AI-generated content, this uncertainty is not merely academic. Contracts built on the assumption that AI output is freely owned by the deployer may be defective. Technology licensing arrangements that purport to assign rights in AI-generated works may convey nothing of value if those works are not protectable. Investors acquiring Ukrainian technology companies should treat unresolved AI IP ownership as a material due diligence issue.
To explore how your AI-driven content strategy interacts with Ukrainian intellectual property rules, contact us at info@ferrazwhitmore.com.
Ownership chains, technology licensing, and the due diligence challenge
Ownership of AI-generated works in Ukraine is not simply a doctrinal puzzle. It is a transactional risk. When a business builds its commercial model on content generated by AI systems, every downstream agreement. technology licensing deals. Content distribution contracts, investor representations. rests on an implicit assumption that the company owns or controls what it is licensing or representing.
That assumption warrants careful examination. Ukrainian intellectual property legislation recognises several mechanisms for allocating rights in works created in an employment or commission context. Works created by employees in the course of their duties vest in the employer under employment legislation, subject to contract. Commissioned works may vest in the commissioner if the agreement is properly drafted. These rules were designed for human creators. Applying them to AI output requires an intermediate step: first establishing that human authorship exists at all, and only then determining where that authorship sits in the ownership chain.
The practical implication for technology licensing is significant. A licensor that cannot establish protectable authorship in its AI-generated content is licensing something that may have no legal substance. The licensee receives no effective exclusivity. If a competitor reproduces the same content, there is nothing to enforce. For companies in the digital services sector, where AI-generated assets may constitute the bulk of commercial value, this gap can be commercially devastating.
Due diligence in acquisitions of Ukrainian technology businesses should therefore include a systematic review of how AI-generated content is produced and documented. Specifically, the review should assess whether human creative decisions are recorded at each stage of the production workflow. It should also evaluate whether third-party training data was licensed or is subject to open-source conditions that permit commercial use. Where gaps are identified, remedial steps – retrospective licensing, content replacement, or contractual risk allocation – should be negotiated before closing.
For businesses whose AI systems were trained on Ukrainian literary, artistic, or musical works, a separate exposure exists. Rights holders in Ukraine have begun paying closer attention to AI training practices. The window for proactive licensing of training data is narrowing. Companies that act now to regularise their training data position will be better placed than those who wait for enforcement action.
Our analysis of the parallel questions arising in the Russian legal context illustrates how CIS jurisdictions are developing divergent doctrinal responses to the same underlying challenge.
Cross-border implications for CIS-connected businesses
Ukraine's intellectual property legislative regime operates within a set of international obligations. Ukraine is a party to the Berne Convention for the Protection of Literary and Artistic Works and to the WIPO Copyright Treaty. These instruments do not require member states to protect AI-generated works. They do, however, set minimum standards for the protection of human-authored works – including works that happen to be created with AI assistance.
For CIS-connected clients, the cross-border dimension adds complexity. A business incorporated in Georgia or Kazakhstan may commission AI-generated content from a Ukrainian developer. The content is then deployed on platforms serving users in multiple jurisdictions. Ownership questions may need to be resolved under the law of the country where protection is sought, not necessarily under Ukrainian law. This means a single AI-generated work could be protectable in one jurisdiction, unprotectable in another, and subject to competing claims in a third.
The EU's AI Act compliance regime introduces a further external pressure. Ukrainian companies with EU market access. and Ukraine's EU association status makes this a substantial and growing category. must consider whether their AI systems and AI-generated outputs will need to meet EU transparency and documentation standards. AI Act compliance is not yet a formal requirement for Ukrainian businesses operating domestically. However, any company that distributes AI-generated content into the EU, or that enters technology licensing arrangements with EU counterparties, will increasingly face contractual demands for AI Act-compatible documentation.
This creates an asymmetry. A Ukrainian AI company serving only domestic clients operates under an uncertain but relatively permissive domestic regime. The same company serving EU clients faces a significantly more demanding compliance environment. Businesses planning international expansion should build AI Act compliance into their product architecture from the outset, rather than retrofitting it as a market-entry condition.
The intellectual property practice for Ukraine at Ferraz & Whitmore assists clients in mapping their cross-border IP exposure and structuring technology licensing agreements that hold under multiple legal regimes.
For CIS-based investors acquiring Ukrainian AI businesses, a secondary risk concerns the portability of IP assets. Ukrainian corporate legislation permits the assignment of intellectual property rights, including rights in software and creative works. But where the asset being assigned is AI-generated content of uncertain protectability, the assignment may convey a contingent or conditional interest rather than a clean title. Representations and warranties in acquisition agreements should address this explicitly, with appropriate indemnity provisions.
Strategic recommendations and the regulatory outlook
Ukrainian legislative bodies have acknowledged the need to address AI and intellectual property in a dedicated statutory instrument. Draft provisions have circulated in policy discussions. The general direction aligns with EU approaches: recognition of human-directed AI output as potentially protectable, transparency obligations for AI systems used in content generation, and some form of accountability regime for AI deployers. However, no comprehensive AI legislation has yet been enacted in Ukraine. The timeline for such legislation remains uncertain given the current geopolitical context.
In the absence of clear statute, strategic protection of AI-generated works in Ukraine must be built on existing doctrinal tools, applied carefully. Several approaches are available.
First, businesses should invest in workflow documentation. Every stage at which a human makes a creative decision in the AI content production process should be recorded. This includes prompt selection, curation of outputs, editorial revision, and arrangement into larger works. The documentation creates an evidentiary record that supports an authorship claim if rights are subsequently challenged.
Second, voluntary registration with the State Intellectual Property Service of Ukraine should be considered for commercially significant AI-assisted works. Registration does not create rights, but it establishes a dated, official record of the registered party's claim. In litigation, that record shifts the burden of proof. For high-value digital services assets, the cost of registration is modest relative to the evidentiary benefit.
Third, technology licensing agreements should be reviewed and, where necessary, restructured. Licences that purport to convey rights in AI-generated content should include representations about the production process. Warranties as to the absence of third-party training data claims. Additionally, indemnity provisions that allocate the risk of future regulatory developments. A bare assignment of "all rights in AI-generated works" is likely insufficient under current Ukrainian doctrine.
Fourth, training data governance deserves immediate attention. Companies that have not audited the provenance of their training data are exposed to claims from Ukrainian rights holders. A systematic audit – identifying what works were used, under what conditions, and whether any outstanding licensing obligations exist – should precede any significant commercial deployment of AI systems trained on Ukrainian content.
Fifth, companies with EU market exposure should begin mapping their AI Act compliance obligations now. The EU's requirements around transparency, human oversight, and documentation of AI systems will increasingly be mirrored in commercial contract demands from EU counterparties. Early compliance investment reduces the cost of market entry later.
The regulatory trajectory in Ukraine points toward greater specificity. Practitioners who advise technology clients expect that Ukrainian intellectual property legislation will, within the next several years, introduce dedicated provisions addressing AI-generated works. The most likely outcome is a model that recognises human-directed AI output as protectable, with disclosure obligations attached to AI-generated content used in commerce. Companies that build their IP strategies on that anticipated baseline will be better positioned than those whose strategies depend on continued legislative silence.
For a tailored strategy on AI intellectual property protection and technology licensing in Ukraine, reach out to info@ferrazwhitmore.com.
Self-assessment checklist for businesses with AI-generated IP in Ukraine
The analytical approach described in this article applies if your business meets one or more of the following conditions. You develop, deploy, or distribute AI systems that produce creative output – text, images, code, audio, or video – in or from Ukraine. You hold technology licensing agreements that include AI-generated content as a licensed asset. You are acquiring or investing in a Ukrainian company whose commercial value depends on AI-generated works. You use AI systems trained on Ukrainian literary, artistic, or musical works.
Before relying on IP ownership claims over AI-generated content in Ukraine, verify the following. First, can you document the human creative decisions made at each stage of the AI production workflow? Second, do your technology licensing agreements address the possibility that AI output may not be protectable under current Ukrainian law? Third, has the provenance of your training data been audited for Ukrainian rights holder exposure? Fourth, do your contracts with EU counterparties anticipate AI Act compliance demands? Fifth, have commercially significant AI-assisted works been voluntarily registered with the State Intellectual Property Service?
If the answer to any of these questions is uncertain, the gap between your current IP position and a defensible one is material. Acting to close that gap before a rights holder files a claim, or before a transaction counterparty identifies the exposure in due diligence, is the more commercially sound path.
For companies whose business models place AI-generated content at the centre of their value proposition. The AI and technology law practice for Ukraine at Ferraz &. Whitmore provides the doctrinal and transactional support needed to build a position that holds.
Frequently asked questions
Q: Can an AI system be recognised as an author of a creative work under Ukrainian law?
A: No. Ukrainian intellectual property legislation ties authorship exclusively to natural persons. An AI system cannot hold authorship rights. The practical question is whether the human who directed, prompted, or trained the AI qualifies as an author. and that analysis turns on the degree of creative contribution the human made to the final output.
Q: How long does it take to register an IP claim over an AI-assisted work in Ukraine, and what does it cost?
A: Copyright in Ukraine arises automatically on creation and does not require registration. Voluntary registration through the State Intellectual Property Service typically takes several weeks to a few months depending on document completeness. Government fees are set by regulation and are modest by international standards; legal fees for preparing and filing a registration package start from a few hundred euros. Registration creates a rebuttable presumption of authorship, which is valuable when AI involvement complicates ownership chains.
Q: A common belief is that training an AI on publicly available Ukrainian works is always lawful – is that correct?
A: This is a widespread misconception. Public availability does not equal permission to reproduce works for AI training purposes. Ukrainian copyright legislation does not contain a broad text-and-data-mining exception comparable to those being introduced in the EU. Rights holders whose works are ingested without authorisation may have viable infringement claims, particularly where the training output competes with the original market for the source work. Engaging a lawyer in Ukraine with cross-border AI experience is advisable before deploying any AI system trained on third-party Ukrainian content.
About Ferraz & Whitmore
Ferraz & Whitmore is an international law firm based in Lisbon, advising business clients across 46 jurisdictions. Our AI and technology law practice covers AI-generated intellectual property, software liability, algorithmic accountability, and technology licensing across CIS and international markets. We work with technology developers, digital services companies, institutional investors, and in-house legal teams who need results-oriented counsel on AI regulation and IP strategy. As an international law firm operating across Ukraine and neighbouring jurisdictions, we bring both civil law depth and common law analytical rigour to matters where the two traditions intersect. Our attorneys have advised on AI Act compliance strategies and cross-border technology licensing transactions spanning both EU and non-EU legal regimes. The firm's Lisbon base provides direct access to EU regulatory developments, while our CIS practice supports clients managing risk across Ukraine, Georgia, Kazakhstan, and related markets. To discuss how AI-generated IP rules in Ukraine apply to your specific situation, 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.