A technology company deploying a generative AI tool in Tashkent produces a body of creative output. designs, written content, software modules – and then discovers that no clear legal owner exists for any of it. The gap between how the product was built and how Uzbek intellectual property legislation defines authorship is not a theoretical problem. It is a commercial one, with direct consequences for licensing revenue, investor due diligence, and enforcement rights.
Uzbekistan's intellectual property legislation, rooted in a civil law tradition, requires a natural person as the author of a protected work. AI-generated works produced without sufficient human creative input fall outside this definition, leaving ownership unresolved. The absence of dedicated AI regulation means that operators, developers, and users currently rely on contract law, software protection rules, and analogical reasoning to fill the gap.
This analysis examines the doctrinal foundations of Uzbek IP law as applied to AI-generated works, reviews emerging interpretive tensions. Identifies the practical gap between statute and enforcement. Additionally, draws out the cross-border and strategic implications for international businesses operating across the CIS region.
Doctrinal foundations: what Uzbek intellectual property law says about authorship
Uzbekistan's intellectual property legislation follows the civil law model that prevails across the CIS. Authorship is a personal right attached to a human creator. The statute does not contemplate a non-human author. This is not a gap unique to Uzbekistan – it reflects the architecture of Berne Convention membership, which all CIS states share. What distinguishes Uzbekistan is the relatively early stage of its AI regulatory development and the limited volume of court-tested doctrine in this specific area.
Under Uzbek intellectual property legislation, a work receives protection when it is original and expressed in a tangible form. Originality, as practitioners in Uzbekistan interpret it, requires that a natural person's intellectual effort shaped the final expression. A work produced entirely by an algorithm, without human selection, arrangement, or editorial judgment, does not satisfy this requirement. The consequence is stark: the work enters a legal space that the statute does not address, neither protected property nor clearly in the public domain.
Software receives a specific category of protection under Uzbek intellectual property legislation. Computer programs are treated as literary works. This creates a layered question for AI systems: the underlying model may be protected as software, but the outputs it generates are assessed separately under the authorship test. A developer who writes code that produces text or images does not automatically hold copyright in every output the code generates. The link between the protected tool and the unprotected output is one of the most commercially significant doctrinal gaps in the current system.
Database protection offers a partial alternative. Where an AI system organises or compiles data in a structured form, the resulting database may attract protection under the suverenlik (sovereign legislative) framework governing related rights. This is distinct from authorship-based copyright. It does not require human creativity in the individual outputs, only human investment in the structure. For businesses whose AI tools generate structured datasets, this route deserves attention. It is, however, narrower than full copyright protection and does not cover standalone creative outputs such as images, music, or articles.
Practitioners in Uzbekistan note that the civil law tradition creates a further complication: moral rights attach to the author personally and are not transferable. Even where a human author can be identified for an AI-assisted work, the moral rights question. who holds the right of integrity. The right of attribution. adds a layer of complexity that standard technology licensing agreements rarely address.
Competing interpretations and the gap between statute and practice
In the absence of dedicated AI legislation, Uzbek courts and practitioners have applied existing intellectual property rules by analogy. Several interpretive approaches have emerged, and they do not yet point in a single direction.
The first approach treats the human operator as the author. Under this view, the person who configured, prompted, or directed the AI system exercised sufficient creative judgment to qualify as the author of the output. The AI is treated as a sophisticated tool – comparable to a camera or a word processor – and the user's creative choices determine the protected expression. This approach preserves the human-authorship requirement and extends protection to commercially valuable AI-assisted content. Its weakness is that it stretches when the AI's contribution dominates: if the output would be substantially the same regardless of how the user framed the prompt, the user's creative contribution becomes difficult to sustain.
The second approach focuses on the developer. Under this model, the entity that trained the model and defined its parameters made the formative creative choices. The operator and user merely release the results of those choices. This approach has appeal in theory but is operationally difficult. Developers are typically not parties to the end-use relationship. They rarely intend to hold copyright in every output their model generates across thousands of deployments. Technology licensing arrangements almost universally disclaim any ownership interest by the developer in user-generated outputs.
A third position, held by a minority of commentators in the region, is that purely AI-generated works fall into the public domain immediately on creation. No registration is needed to remove protection; it simply never attaches. This view has the virtue of doctrinal consistency with the human-authorship requirement. It creates significant commercial problems, however. A business that invests in AI-generated content and cannot protect that content against copying has lost a central part of its return on that investment. The lost-opportunity cost is direct and measurable.
The gap between statute and practice emerges most clearly at the enforcement stage. Even where a business structures its agreements to assign all rights in AI-assisted outputs to a corporate entity. relying on employment law provisions or work-for-hire analogies. enforcement against an infringer requires the court to accept that a protectable right exists in the first place. If the defendant challenges the existence of copyright in an AI-generated work, the plaintiff bears the burden of demonstrating a qualifying human creative contribution. This is a burden that many businesses do not anticipate when they deploy AI tools, and it has direct consequences for algorithmic accountability structures and software liability allocations in commercial contracts.
For a business operating between Uzbekistan and a jurisdiction with more developed AI IP doctrine – the European Union, the United Kingdom, or Singapore – this doctrinal gap creates an asymmetry. The same work may attract full protection in the home jurisdiction and uncertain or no protection in Uzbekistan. Cross-border licensing agreements must account for this difference explicitly. A licensee in Tashkent who receives a licence for AI-generated content may be receiving a licence to something the licensor cannot actually own under Uzbek law. The downstream risk to both parties is underappreciated.
For a detailed comparison of how these questions are being addressed in an adjacent CIS jurisdiction. See our analysis of AI-generated works and intellectual property in Russia. There, a somewhat more developed body of commentary has begun to form.
To explore how these doctrinal questions affect your technology licensing strategy in Uzbekistan, contact us at info@ferrazwhitmore.com.
AI Act compliance, algorithmic accountability, and digital services in the Uzbek context
Uzbekistan does not yet have a dedicated AI Act or comprehensive AI regulation. The country's digital transformation agenda is active: legislative initiatives on digital services, e-commerce, and data protection have advanced in recent years. However, a statutory framework specifically governing AI systems – their development, deployment, and outputs – has not been enacted at the time of this writing.
This means that AI Act compliance, in the sense applicable in the European Union, has no direct Uzbek counterpart. International businesses subject to EU AI Act obligations who also operate in Uzbekistan must manage two distinct compliance environments. The EU framework imposes risk classification, transparency, and documentation requirements on AI systems. Uzbek law imposes none of these specifically. The practical effect is that a business deploying an EU-compliant AI system in Uzbekistan benefits from its compliance posture internally but cannot rely on it as a legal shield in Uzbek proceedings.
Algorithmic accountability is an emerging theme in Uzbekistan's broader digital services policy. Regulatory guidance on digital services has addressed questions of platform liability and content moderation. These provisions touch on, without fully resolving, the question of who is responsible when an automated system produces harmful or legally problematic output. Software liability in this context follows general civil law principles: liability attaches to the legal person who deployed or operated the system, not to the system itself.
Technology licensing agreements in Uzbekistan must therefore be drafted with this liability allocation clearly in mind. A standard international software licence that relies on the licensor's EU AI Act compliance documentation as the basis for liability limits will not translate cleanly into Uzbek proceedings. The indemnity and limitation-of-liability clauses must be assessed against Uzbek civil legislation, which has its own approach to consequential loss and limitation clauses. Practitioners advising on technology licensing in this market consistently flag this as an area where international templates require local adaptation.
The regulatory trajectory suggests that Uzbekistan will introduce more specific AI and digital services rules over the medium term. The country has joined international digital economy initiatives and has signalled legislative intent in this direction. Businesses that build their IP and liability structures now – before that legislation arrives – will be better positioned to adapt. Those that wait until a formal AI regulatory regime is in place risk finding that their existing agreements and ownership structures do not map onto the new requirements.
Cross-border implications for CIS clients and international operators
The CIS region presents a varied picture on AI and intellectual property. Several jurisdictions are further along in their engagement with AI-specific IP questions. Uzbekistan's approach is influenced by these regional developments, though it retains its own legislative pace and judicial culture.
For international businesses structuring their AI operations across CIS markets, the Uzbek position creates specific challenges. Technology licensing arrangements that work in one CIS jurisdiction may not achieve the same result in another. An IP holding structure established in a more permissive jurisdiction does not automatically confer enforceable rights in Uzbekistan. The cross-border question is not merely academic: CIS businesses routinely use regional holding companies to centralise IP ownership. Additionally. Those structures must be tested against the authorship and protection rules of each jurisdiction where the IP will be used or enforced.
For the intellectual property registration and protection strategy that underpins any cross-border AI deployment, the foundation matters. Our detailed review of intellectual property law in Uzbekistan sets out the registration procedures, available rights, and enforcement mechanisms in the current statutory environment.
Enforcement across CIS borders raises the recognition question directly. A judgment from a Uzbek court recognising or denying IP protection for AI-generated content may be difficult to enforce in a third jurisdiction, and vice versa. The Minsk Convention framework, which governs mutual legal assistance among CIS states, covers civil and commercial matters but does not address AI-specific IP questions. Parties relying on arbitration clauses in technology agreements gain more flexibility: major arbitral institutions' rules allow the arbitral tribunal to determine applicable law. This may provide a more predictable route to resolution than domestic courts in any single CIS jurisdiction.
A non-obvious risk that many international operators overlook is the interaction between AI-generated content and Uzbekistan's content regulation rules. Digital services operating in the Uzbek market must comply with local content rules. Where AI-generated content is published or distributed in Uzbekistan, the operator bears liability for that content under Uzbek digital services legislation – regardless of whether the content is protected by copyright or not. The absence of IP protection does not reduce regulatory exposure. It simply means the operator has no copyright-based tool to control how others use the same content.
For businesses licensing AI-generated content to Uzbek counterparties, this asymmetry is commercially significant. The licensor may be unable to prevent the licensee from reproducing the content freely if Uzbek courts decline to recognise copyright. The licensor retains only contractual remedies – a breach of contract claim against the specific licensee – rather than the broader property-based remedies that copyright would provide. Contract drafters must therefore include explicit reproduction restrictions, penalty clauses, and audit rights that do not depend on the existence of underlying copyright.
To discuss how cross-border IP and AI strategies should be structured for operations in Uzbekistan and across CIS markets, contact us at info@ferrazwhitmore.com.
Strategic recommendations for businesses operating in Uzbekistan
The doctrinal uncertainty described above does not leave businesses without options. A structured approach to IP protection and contract design can significantly reduce exposure, even before Uzbekistan enacts dedicated AI legislation.
The first priority is to document human creative contribution at every stage of AI-assisted production. This is the single most effective measure for preserving copyright eligibility under current Uzbek intellectual property legislation. Documentation should record who made selection and editorial decisions, what prompts or parameters were used, and how the final output was reviewed and modified by a human. This record does not guarantee protection – courts retain discretion – but it provides the evidentiary foundation for an authorship argument.
The second priority is to structure technology licensing agreements for the Uzbek market specifically. International template agreements routinely fail on at least three points when applied to Uzbek AI deployments: they do not address the local authorship gap. They rely on EU or US liability limitation approaches that may not be enforceable under Uzbek civil legislation. Additionally, they omit explicit contractual protections that substitute for the copyright remedies that may be unavailable. A law firm in Uzbekistan with technology practice experience can identify the specific adaptations required for each of these points.
Third, businesses should consider whether database protection, rather than copyright, is the more reliable route for structuring their AI-generated assets. Where outputs are structured and compiled in a systematic way, database-related rights may provide a more defensible protection argument than authorship-based copyright. This requires planning at the product architecture stage, not after deployment.
Fourth, for higher-value AI-generated content, arbitration clauses with a neutral seat and governing law provide a more predictable enforcement path than reliance on Uzbek domestic courts in novel IP disputes. Selecting a seat in a jurisdiction with developed AI IP doctrine – and specifying the applicable law of that jurisdiction – transfers the interpretive question to a tribunal better equipped to address it.
Self-assessment checkpoints before deploying AI tools in Uzbekistan include:
- Can the business document specific human creative choices in every output category it intends to protect?
- Do existing technology licensing agreements contain Uzbek-law-specific liability and reproduction provisions?
- Has the IP holding structure been tested against Uzbek authorship rules, not only the rules of the holding jurisdiction?
- Are AI-generated works that will be published in Uzbekistan assessed for local content regulation compliance, separate from the IP protection question?
- Does the dispute resolution clause in key technology agreements provide a realistic enforcement path for AI IP disputes?
Businesses that address these points proactively retain significant strategic options. Those that deploy AI tools in Uzbekistan under standard international templates without local adaptation are forfeiting those options – often without realising it until a dispute arises.
The full range of AI and technology law services available to businesses entering or expanding in the Uzbek market is set out at AI and technology law in Uzbekistan.
Outlook: what to monitor as Uzbek AI law develops
Uzbekistan's legislative agenda in the digital and technology space is moving. Several developments merit close attention by businesses with existing or planned AI deployments in the country.
The first is the anticipated development of AI-specific legislation or regulatory guidance. Regional signals – including activity at the CIS intergovernmental level and Uzbekistan's participation in digital economy frameworks – suggest that some form of AI regulatory instrument is likely within the medium term. When it arrives, it will almost certainly address the authorship gap, though the direction it takes is not yet predictable. Businesses that have structured their operations under current rules should monitor legislative drafts and ensure their agreements contain adaptation clauses.
The second development to watch is the emerging body of court practice in digital services and software disputes. Even without AI-specific legislation, courts addressing software liability and digital content disputes are building interpretive habits that will shape how AI IP questions are decided when they arrive in full. Practitioners note that Uzbek courts are increasingly receptive to foreign legal commentary and comparative analysis in novel technology cases. This opens space for well-prepared arguments drawing on international doctrine.
Third, the interaction between Uzbekistan's data protection legislation and AI systems is becoming commercially relevant. AI models trained on personal data processed in Uzbekistan must comply with local data protection rules. Where training data includes content generated by or about Uzbek residents, the data protection and IP questions overlap. A work generated by an AI model that was trained on unlicensed third-party content raises both a potential IP infringement claim and, where personal data was involved, a data protection compliance issue. These overlapping exposures are best assessed together rather than in isolation.
Finally, the regional dimension will continue to matter. As CIS jurisdictions develop their own AI IP rules at different speeds and in different directions, the pressure for some degree of regional harmonisation will grow. Businesses that build flexible, jurisdiction-aware IP structures now will be better placed to adapt to that harmonisation. or to exploit the arbitrage opportunities that persisting differences create. than those locked into rigid frameworks designed for a single market.
Frequently asked questions
Q: Can an AI system be named as the author of a work under Uzbek intellectual property legislation?
A: No. Uzbekistan's intellectual property legislation ties authorship to natural persons. An AI system cannot hold authorship rights. The question of who – the developer, the operator, or the user – qualifies as the human author of an AI-assisted work remains doctrinally unresolved, and no definitive court guidance has yet emerged.
Q: How long does it take to register a copyright or software right in Uzbekistan, and what does it cost?
A: Copyright in Uzbekistan arises automatically on creation and does not require registration. Voluntary deposit with the state body typically takes several weeks and involves moderate administrative fees. Software registration as a database or computer program follows a similar timeline. For foreign applicants, additional notarisation and legalisation steps add time – often four to eight weeks in total.
Q: Is it a misconception that foreign copyright in AI-generated content is automatically protected in Uzbekistan?
A: Yes. A common misconception is that Berne Convention membership guarantees seamless protection for AI-generated works across all member states. In practice, Uzbekistan's courts and regulatory bodies apply domestic authorship requirements first. If a work lacks a qualifying human author under Uzbek law, protection may be denied regardless of the status accorded in the country of origin. Engaging a lawyer in Uzbekistan with cross-border IP experience is essential before assuming foreign rights will be recognised.
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 technology licensing across CIS and international markets. We work with technology companies, institutional investors, and in-house legal teams who need results-oriented counsel on AI-generated works, digital services regulation, algorithmic accountability, and software liability matters in high-growth and emerging markets. As an international law firm advising on Uzbekistan, we bring direct experience before arbitral bodies including the ICC and SIAC. Additionally. Our AI and technology practice spans civil law and common law systems across the CIS, Asia-Pacific, and European jurisdictions. The firm's Lisbon base provides direct access to EU regulatory developments – including AI Act compliance requirements – while our CIS network supports on-the-ground advisory in Tashkent and across the region. To discuss your AI IP strategy in Uzbekistan, 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.