A technology company deploys a generative AI system to produce marketing copy, visual assets, and product documentation for the Russian market. The output is commercially valuable. Then a competitor reproduces it without licence. The company's legal team reaches for intellectual property legislation – and discovers that the question of who, if anyone, holds copyright in an AI-generated work in Russia has no settled answer. The gap between what the statute says and what courts are willing to hold is wide, contested, and growing.
Under Russian intellectual property legislation, copyright protection attaches only to works created through the creative effort of a natural person. AI-generated outputs currently fall outside that definition unless a human contributor can demonstrate sufficient creative involvement. No dedicated statutory regime for AI-generated works yet exists in Russia, leaving owners to rely on doctrinal arguments, contractual arrangements, and adjacent rights such as software protection or trade secret rules.
This analysis examines the doctrinal foundations of Russian copyright doctrine as they apply to AI-generated content, maps the gap between statute and emerging court practice. Considers cross-border implications for CIS clients. Additionally, sets out strategic recommendations for businesses that cannot afford to leave their AI-generated assets unprotected.
Doctrinal foundations: creativity, authorship, and the human requirement
Russian intellectual property legislation is built on a civil law model of authorship. Protection arises when a natural person expresses creative effort in producing an original result. The legislation does not define "creative effort" with precision. Courts and scholars have debated for decades whether minimal originality suffices or whether something more substantial is required.
In practice, courts in Russia apply a relatively low threshold for human-authored works. A modest degree of individual choice – in composition, selection, or arrangement – has historically been enough. The decisive question for AI-generated outputs is whether that choice was exercised by a person or by an algorithm operating on probabilistic inference.
The legislation draws a clear line: legal entities and software systems cannot be authors. A company may own copyright by assignment or as the employer of a human author. But if no human author exists – because the work was produced end-to-end by an AI model responding to a brief – the chain of assignment has no starting point. The work, in strict doctrinal terms, enters the public domain at the moment of creation.
This position is consistent with the general approach taken in most civil law systems, including those of other CIS states. It also aligns with the position that courts in several common law jurisdictions have taken when confronted with analogous questions. The convergence is not coincidental: it reflects a shared doctrinal commitment to the idea that intellectual property rights exist to incentivise human creativity, not machine output.
Practitioners in Russia note, however, that the human authorship rule was formulated before generative AI existed as a commercial reality. Applying it to outputs from large language models or image generation systems produces results that many in the technology sector regard as commercially unworkable. A significant share of the content produced by Russian technology companies today involves some degree of AI generation. Leaving that content unprotected creates a structural vulnerability that competitors are already beginning to exploit.
Competing interpretations: where courts and commentators diverge
The absence of dedicated AI legislation in Russia has left courts to resolve disputes by analogy and doctrinal extension. The results are not uniform. Several distinct interpretive positions have emerged, and understanding where they diverge is essential for any business assessing risk.
The first position holds that AI-generated works are simply unprotected. On this view, the statute is clear. No human author means no copyright. The work is available for anyone to use. Courts applying this line reason that expanding protection by judicial interpretation would require the legislature to act first. This is the most conservative position, and it is also the most legally coherent given the current text of Russian intellectual property legislation.
The second position attributes authorship to the human operator who directed the AI. Proponents argue that selecting prompts, curating outputs, and integrating results into a commercial context constitutes sufficient creative effort. The AI is treated as a sophisticated tool – analogous to a camera or a word processor – rather than an independent creative agent. This view has found some support in academic commentary and in at least one line of lower court reasoning, though it has not been definitively endorsed at the appellate level.
The third position – the most commercially convenient but the least doctrinally secure – treats the deploying legal entity as the rights-holder under a sui generis rule. Advocates point to existing provisions in Russian legislation that grant neighbouring rights to producers of phonograms and databases without requiring a human creative act. They argue that similar logic should apply to AI outputs. The Intellectual Property Court (Sud po Intellektualnym Pravam), Russia's specialised IP tribunal, has not yet issued a definitive ruling endorsing this position, but it has acknowledged the doctrinal gap in several procedural decisions.
The practical consequence of this divergence is significant. A business that relies on the second position – human-operator authorship – may find that a different court applies the first position and denies protection entirely. The risk is not theoretical. Technology licensing agreements, content distribution contracts, and software liability allocations all depend on knowing who holds the underlying rights. Uncertainty on that question propagates through every downstream commercial arrangement.
For a detailed analysis of how Russian AI law advisory intersects with technology licensing and digital services transactions, see our AI and technology law services in Russia.
The gap between statute and practice: software, databases, and adjacent protection routes
Where primary copyright protection is unavailable, Russian intellectual property legislation offers several adjacent protection mechanisms. Each has distinct conditions, timelines, and limitations. Understanding which route applies – and when it becomes insufficient – is the core of a sound IP strategy for AI-generated content.
Software protection is the most frequently used route. Russian legislation protects computer programs as a species of literary work. The protection attaches to the software itself – the code, the architecture, the training pipeline – rather than to the output the software produces. A company that develops or licences a proprietary AI model can assert rights in that model. It cannot, on current doctrine, automatically assert rights in everything the model generates. This distinction matters enormously in technology licensing negotiations: a licensee who gains access to an AI tool does not thereby acquire rights in the tool's outputs unless the agreement explicitly addresses that question.
Database protection offers a second route for AI systems that compile, organise, or structure information. Russian intellectual property legislation provides a neighbouring right for producers of databases who have made a substantial investment in obtaining, verifying, or presenting the contents. This right is not contingent on creative effort. It protects the investment, not the expression. For AI systems that generate structured datasets, training corpora, or knowledge bases, this neighbouring right may provide a meaningful layer of protection even where copyright is unavailable.
Trade secret protection under Russian commercial legislation is a third option. If the AI-generated output qualifies as commercially valuable information that is kept confidential through reasonable measures, it may attract trade secret status. The protection is not a property right in the traditional sense – it does not prevent independent creation or reverse engineering – but it does provide a basis for claims against misappropriation and unauthorised disclosure. For many technology companies, trade secret rules offer the most practically accessible protection for AI outputs in the short term.
A common mistake among international clients is to treat these adjacent mechanisms as substitutes for copyright. They are not. Software protection covers the model, not the output. Database rights require substantial investment and do not cover single generated items. Trade secret rules require active confidentiality management and fail the moment the output is publicly distributed. Each mechanism covers a specific situation. None provides the broad, default protection that copyright would give if the human authorship requirement were relaxed.
Algorithmic accountability obligations add a further layer of complexity. Russian digital services legislation imposes disclosure and audit obligations on operators of algorithmic recommendation systems. A company that generates and distributes AI-produced content at scale may find that its distribution infrastructure is subject to these obligations independently of any IP question. Software liability for defective AI outputs is a related area where Russian courts are beginning to apply general product liability and digital services rules by analogy, with outcomes that are still unpredictable.
To understand how intellectual property protection strategies interact with broader Russian IP registration and enforcement practice, see our analysis of intellectual property law in Russia.
To receive expert guidance on structuring AI asset protection across Russia and the CIS region, contact us at info@ferrazwhitmore.com.
Cross-border implications for CIS clients and international businesses
The Russian position on AI-generated works does not exist in isolation. Businesses operating across the CIS region face a mosaic of national rules that are each at a different stage of doctrinal development. Understanding how these systems interact – and where Russian law creates particular risks for cross-border transactions – is essential for any operator with assets or revenues in multiple CIS jurisdictions.
Kazakhstan, Uzbekistan, and Belarus each maintain intellectual property legislation derived from the same Soviet-era civil law tradition as Russia. All three apply the human authorship requirement. None has yet enacted dedicated AI-specific copyright rules. The practical result is that a business producing AI-generated content in Russia and distributing it across the CIS region faces the same doctrinal gap in each jurisdiction. Applied through different courts and administrative bodies with different levels of sophistication on technology questions.
The cross-border dimension becomes especially acute in technology licensing transactions. A licensor based outside the CIS who grants rights in AI-generated content to a Russian or Kazakhstani licensee should assume that the licensee's courts may not recognise those rights as valid property. The licence purports to convey something that the receiving jurisdiction's law may treat as being in the public domain. Structuring the agreement to rest on software rights, contractual exclusivity, or trade secret obligations – rather than on copyright in the output – significantly reduces this exposure.
For businesses with EU operations, the interaction with EU AI Act compliance requirements creates an additional planning challenge. The EU AI Act imposes obligations on providers and deployers of AI systems that are placed on the EU market or affect persons in the EU. A company that develops AI-generated content in Russia and exports it to EU customers must consider both Russian IP rules governing the creation of that content and EU AI Act compliance obligations governing its deployment. These are parallel, non-overlapping regimes. Meeting one does not satisfy the other. In practice, the gap most frequently appears in transparency and labelling obligations: EU rules may require disclosure that content is AI-generated. While Russian digital services rules impose different and sometimes conflicting disclosure obligations for the same material.
Sanctions and trade control rules add a further layer for businesses with any Western nexus. Technology licensing agreements involving Russian AI systems or datasets may engage export control regimes in the EU, the UK, or the United States. The intersection of these controls with IP questions – particularly where training data or model weights are involved – requires careful analysis before any cross-border arrangement is concluded.
CIS clients investing in AI development for regional distribution should also consider the choice of governing law and dispute resolution forum with particular care. Arbitration clauses designating neutral seats outside Russia. the Mezhdunarodny Kommerchesky Arbitrazhny Sud (International Commercial Arbitration Court. ICAC) in Moscow handles a large volume of IP disputes involving Russian parties. provide a more predictable enforcement environment for IP claims than domestic court litigation in jurisdictions where AI-generated works doctrine is still unsettled.
For a comparative perspective on how analogous questions are handled in the Kazakhstani context, see our deep analysis of AI-generated works and intellectual property in Kazakhstan.
To explore how cross-border IP structuring strategies apply to your specific situation in Russia and the CIS, reach out to info@ferrazwhitmore.com.
Strategic recommendations and the regulatory outlook
The current doctrinal uncertainty in Russia does not mean that businesses must accept unprotected AI-generated assets. It means that protection requires deliberate structuring rather than reliance on default rules. Several strategies, applied in combination, materially reduce the risk of losing rights in commercially valuable AI outputs.
The first priority is documentation. Every AI-generated work that has commercial value should be accompanied by contemporaneous records of the human decisions involved in its creation: the selection of the AI system. The design of prompts, the curation of outputs. Additionally, the editorial choices made before final use. This documentation does not guarantee copyright protection. It does, however, provide the factual foundation for the human-operator-authorship argument if that argument is needed in litigation or in a licensing negotiation.
The second priority is layered protection. Relying on a single legal mechanism is insufficient. A sound strategy combines software protection for the AI model, database neighbouring rights for structured outputs, trade secret rules for confidential materials, and contractual exclusivity provisions in every downstream agreement. The layers are not redundant – each covers a distinct vulnerability that the others do not address.
The third priority is contract drafting. Technology licensing agreements should explicitly address the ownership of AI-generated outputs. This includes: specifying which party holds any rights that may arise in outputs; allocating the risk of a court finding that no rights exist; and providing fallback protections through confidentiality and non-compete obligations. Many standard form licences used in international technology transactions were drafted before generative AI became commercially significant. They do not address these questions adequately. Adapting them to the Russian context requires attention to both the IP gap and the specific requirements of Russian commercial legislation governing digital services.
The fourth priority is monitoring regulatory developments. Russia has signalled an intent to develop dedicated AI legislation. Presidential strategy documents and proposals circulating within the legislative process suggest that an AI regulatory regime is a medium-term priority. The likely direction of reform. based on the trajectory of Russian digital services legislation and the academic commentary that has influenced prior technology law reforms. is toward some form of sui generis protection for AI outputs. Most probably vested in the deploying entity rather than attributed to a fictional human author. Businesses that structure their IP arrangements now to anticipate this direction will be better positioned when the legislation arrives.
The fifth priority is jurisdictional awareness. For businesses with CIS-wide operations, the choice of where to hold IP rights. There. To register them. Additionally. This law governs disputes over them is a strategic decision that should be made with full awareness of the doctrinal positions in each relevant jurisdiction. Holding AI-related IP in a jurisdiction with more developed protection – and licensing into Russia and other CIS states from that holding structure – remains the most reliable protection strategy pending legislative reform.
Frequently asked questions
Q: Who owns copyright in an AI-generated work under Russian law today?
A: Russian intellectual property legislation currently attributes authorship only to natural persons who exercise creative effort. Where a work is generated autonomously by an AI system, with no identifiable human creative input, no copyright automatically vests. The individual or entity that deployed the AI may seek protection through contractual arrangements or by demonstrating sufficient creative contribution, but the position remains unsettled and contested in practice.
Q: How long does it take to register an IP right in Russia, and what does it cost?
A: Copyright in Russia arises automatically upon creation and requires no formal registration. Registration of patents, trademarks, and software with Rospatent (the Federal Service for Intellectual Property) typically takes between several months and two years depending on the type of right and examination load. Government fees are set by regulatory tariff schedules and vary by right type and applicant category. Legal fees for preparation and prosecution add further cost. International applicants should factor in additional time for document translation and legalisation requirements.
Q: Is there a common misconception that EU AI Act compliance automatically satisfies Russian AI regulation?
A: This is a frequent misconception among international technology companies. AI Act compliance addresses EU-specific risk classification, transparency, and conformity assessment obligations. Russia has developed a separate regulatory trajectory for digital services and algorithmic accountability, anchored in domestic legislation and presidential strategy documents. Compliance with one regime does not satisfy the other. Businesses operating across both markets must conduct independent gap analyses for each jurisdiction. Engaging a lawyer in Russia with cross-border AI experience is advisable before launching products in both markets 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. Additionally, technology licensing. including for clients operating in Russia and the wider CIS region. We work with international technology companies, institutional investors, and in-house legal teams who require results-oriented counsel when legal doctrine and commercial reality diverge. Our AI and technology law practice covers regulatory compliance, IP structuring, software liability, and digital services obligations across both civil law and common law systems. The firm's Lisbon base provides direct access to EU regulatory structures, while our CIS practice supports clients before Russian courts and arbitral bodies including the ICAC. As a law firm in Russia advisory practice, we assist clients in designing protection strategies that function across multiple jurisdictions simultaneously. To discuss how current Russian IP doctrine affects your AI-generated assets, 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.