AI-generated inventions Spain is now one of the most pressing questions for founders, in-house counsel, HR teams and investors operating in the Spanish market, and the stakes rise sharply in 2026. With the EU AI Act being phased in, with major obligations applying across 2025 and 2026, companies face new transparency and record-keeping obligations that ripple directly into how ownership of AI outputs is licensed, warranted and transferred. The core legal position remains clear: Spanish and European patent practice requires a natural person as inventor, so ownership is ultimately determined by employment law, contract drafting and commercial strategy rather than by the machine itself.
This guide explains how Spanish law allocates rights, reconciles patentability with inventorship doctrine, and sets out practical clauses, due diligence checklists and negotiation tactics to secure your position.
Intent of this guide: to give founders, in-house counsel, HR and investors in Spain clear, actionable guidance on who owns AI‑generated inventions, both patentable inventions and other IP, how Spanish law views inventorship, and the contractual and practical steps needed to secure rights under the EU AI Act regime.
Disclaimer: This article is general information, not legal advice. Specific situations should be reviewed with qualified local counsel.
The short answer is that machines do not own inventions and cannot, under current Spanish and European practice, be named as inventors. Patent systems require a natural person inventor, which means the human beings who conceive, direct or meaningfully contribute to an invention remain the starting point for ownership analysis. Where AI tools assist that process, the question becomes who among the humans and legal entities involved holds the resulting rights.
In practice, ownership is decided by a combination of employment law, contractual allocation and the nature of the protection sought. For employees, Spanish labour rules and company-invention clauses typically vest rights in the employer where the invention arises within the scope of employment. For contractors, consultants and freelancers, the default position is far less favourable to the commissioning party, so written assignment clauses are essential. For startups and investors, the choice between patent and trade secret protection, combined with robust IP schedules, drives both value and transferability.
The EU AI Act does not rewrite ownership rules, but its phased application adds compliance, transparency and documentation obligations that shape licensing, diligence and contractual warranties. Getting ownership of AI outputs in Spain right therefore now means aligning IP strategy with regulatory compliance from the outset.
Several distinct legal regimes intersect when you analyse AI-generated inventions Spain. Understanding how each regime treats machine-assisted creation, and how they interact, is the foundation for any ownership strategy.
Spanish patent applications are administered by the Oficina Española de Patentes y Marcas (OEPM) under the Spanish Patents Act (Ley 24/2015, de Patentes), and substantive patent practice in Spain aligns closely with the European Patent Office (EPO) on the fundamentals of inventorship and patentability. An invention must be new, involve an inventive step and be susceptible of industrial application. Crucially, both the OEPM and the EPO proceed on the basis that an inventor must be a natural person. The EPO Guidelines for Examination treat inventorship as a designation of a human being who contributed to the inventive concept.
This has direct consequences for AI-generated inventions. Even where an AI system plays a substantial role in generating a technical solution, the application must name one or more human inventors, typically the researchers or engineers who designed, trained, prompted or directed the system in a way that led to the invention. The machine cannot be designated, and an application that attempts to name an AI alone faces refusal.
For outputs that are not inventions in the patent sense, software, text, images, designs, copyright and related rights are the relevant regime. In Spain, copyright is governed by the consolidated text of the Intellectual Property Law (Texto Refundido de la Ley de Propiedad Intelectual, approved by Real Decreto Legislativo 1/1996), published in the Boletín Oficial del Estado (BOE). European copyright doctrine, reflected in guidance from the European Union Intellectual Property Office (EUIPO) and in international policy analysis from the World Intellectual Property Organization (WIPO), generally requires a human author exercising creative choices. Purely autonomous AI output without meaningful human authorship sits in an uncertain zone and may not attract classic copyright protection.
AI copyright in Spain therefore depends heavily on demonstrable human creative input, which affects both ownership and enforceability.
Trade secret protection is frequently the most practical route for AI models, training pipelines, weights and many AI-generated outputs. Spain protects trade secrets through its Trade Secrets Act (Ley 1/2019, de Secretos Empresariales), which implements the EU Trade Secrets Directive and is published in the BOE. To qualify, information must be secret, have commercial value because it is secret, and be subject to reasonable steps to keep it confidential.
Contracts sit above all of these regimes. Whatever the default statutory position, well-drafted employment agreements, service contracts, assignments and licences can allocate ownership and usage rights between the parties. Because AI-generated inventions Spain often involve multiple contributors, employees, contractors, model vendors and data providers, the contractual layer is where ownership is ultimately won or lost. Data protection obligations under the General Data Protection Regulation and Spain’s Organic Law 3/2018 (LOPDGDD), supervised by the Agencia Española de Protección de Datos (AEPD), also bear on this layer: unlawful use of personal data in training can create liabilities that undermine the commercial value of the resulting IP.
The question of whether an AI can be named as inventor has become a defining test case for AI-generated inventions Spain and across Europe. The analysis blends international precedent, Spanish administrative practice and the practical mechanics of naming inventors and securing assignments.
The most influential series of decisions concerns the “DABUS” applications, in which an applicant sought to designate an AI system as the inventor of two inventions. These applications were tested across multiple jurisdictions. Before the EPO, the applications were refused on the ground that the designated inventor must be a human being with legal capacity; an AI system does not meet that requirement. Courts and patent offices in other major jurisdictions, including the United Kingdom and the United States, reached broadly consistent conclusions that an inventor must be a natural person under their respective laws.
The practical lesson from DABUS is that, under current frameworks, there is no route to AI-only inventorship. Where an AI system contributes to an invention, the applicant must identify the human beings whose intellectual contribution justifies their designation as inventors. This is not a mere formality; a false or defective inventorship designation can expose a patent to challenge.
Spanish practice, administered by the OEPM and informed by harmonised European standards, follows the same logic. The likely outcome for any attempt to name an AI as inventor before the OEPM is refusal, consistent with the EPO position. Industry observers expect Spanish courts to align with the European consensus that inventorship presupposes a natural person. Until legislators act at EU or national level, the prudent assumption for AI inventorship in Spain is that a human inventor is mandatory.
That does not mean AI-assisted inventions are unpatentable. It means the human contributors must be correctly identified. Where a team uses AI as a sophisticated tool, the researchers who frame the problem, design the training, select the data, interpret the outputs and reduce the invention to practice are generally the candidates for inventorship.
For applicants, the practical steps are straightforward but important. First, identify the natural persons who made an intellectual contribution to the inventive concept and name them accurately. Second, ensure there is a clean chain of title from those inventors to the entity that will own the patent, typically through employment provisions or express assignment. Third, document the role of the AI tool and the human contributions contemporaneously, so that inventorship can be defended if challenged.
Ownership of AI outputs in Spain therefore turns on assignment mechanics. A named inventor is not automatically the owner; ownership flows from statute and contract. The risk to avoid is a patent that is granted but vulnerable because the chain of title from human inventors to the company is incomplete, or because inventorship was mis-designated to include or exclude the wrong people.
The single most consequential distinction in AI-generated inventions Spain is between work produced by employees and work produced by contractors. The default legal positions differ markedly, and the contractual response must differ accordingly.
Spanish law allocates inventions created by employees primarily through the employee-invention provisions of the Patents Act (Ley 24/2015), read alongside the Estatuto de los Trabajadores. The core principle is that inventions developed by an employee within the scope of their employment, particularly where the role involves inventive or research activity and the employer provides the means, generally belong to the employer. This is often described as the regime for invenciones laborales (employee or service inventions).
Several categories exist in practice: inventions made in performance of an employment contract expressly directed at inventive activity tend to belong to the employer; inventions made by an employee that draw substantially on the employer’s knowledge, means or resources may give the employer a right to claim ownership or a licence, subject to fair compensation; and inventions wholly unconnected to the employment typically remain with the employee. Because these boundaries can be contested, best practice is to include a clear company-invention and IP assignment clause in every relevant employment contract, confirming that inventions and AI outputs created in the course of duties, including those produced with AI tools, vest in the employer, with any statutory compensation addressed expressly.
A high-level employment clause might confirm that all inventions, works and other IP conceived or developed by the employee in the course of employment, whether created with or without the assistance of AI systems, are assigned to the employer upon creation, and that the employee will execute any documents needed to perfect that assignment. This secures employer ownership of AI in Spain and reduces the risk of later disputes.
The position for contractors and service providers is fundamentally different and far riskier for the commissioning party. There is no automatic transfer of IP from an independent contractor to the client merely because the client paid for the work. Absent an express written assignment, the contractor may retain significant rights in the deliverables, including AI models, code and outputs.
For contractor AI ownership, the essential contractual response is an express, present assignment of all IP in the deliverables to the client, coupled with a waiver of moral rights to the extent permissible under Spanish law, warranties of originality and non-infringement, and an obligation to procure assignments from any sub-contractors or personnel. Note that under Spanish copyright law moral rights are, in principle, inalienable, so drafting should reflect the practical limits on waiving them. A model approach provides that the contractor assigns to the client all present and future economic IP rights in the deliverables and any AI-generated outputs created under the agreement, and that the contractor will not assert any retained rights that would impede the client’s use.
Edge cases demand particular care. Consultants engaged on short mandates, freelancers sourced through platforms, and contributors drawing on open-source components can all introduce ownership and licensing complications. Open-source components may carry copyleft obligations that affect how AI outputs can be commercialised. Freelancers may reuse code or models across clients. Volunteer or community contributors may have contributed without any written assignment at all.
The recommended clause language for these scenarios combines express assignment, disclosure obligations requiring the contributor to identify any third-party or open-source materials incorporated, and representations that no incorporated materials impose restrictions inconsistent with the client’s intended use. For AI models specifically, the contract should address ownership of the model, the weights, the training data rights and the outputs separately, because these can belong to different parties under different terms.
Choosing how to protect AI outputs is a commercial as well as a legal decision. For many AI assets, the realistic choice is between patent protection and trade secret protection, with contract assignment underpinning both.
Patents offer strong, enforceable exclusivity but require public disclosure and come with cost and time. Trade secrets avoid disclosure and can be indefinite, but offer no protection against independent development or reverse engineering and depend entirely on maintaining secrecy. The table below compares the principal routes for AI-generated inventions Spain.
| Feature | Patent | Trade Secret | Contract Assignment |
|---|---|---|---|
| Protection scope | Exclusive rights over the claimed invention | Protects confidential information with commercial value | Allocates whatever rights exist between parties |
| Public disclosure | Required, the invention is published | None, secrecy must be maintained | None, terms are private to the parties |
| Term | Limited statutory term from filing (generally up to 20 years, subject to renewal fees) | Indefinite while secrecy is preserved | As agreed; typically perpetual for assigned IP |
| Enforcement | Infringement actions; strong remedies | Misappropriation claims; harder to prove | Breach of contract between parties only |
| Cost / time | Higher cost; examination takes time | Lower upfront cost; ongoing security cost | Low cost; immediate effect |
| Suitability for AI outputs | Good for defined technical inventions with human inventors | Strong for models, weights and pipelines | Essential foundation for all AI assets |
| Effect on transferability / licensing | Clear, registrable asset; easy to licence | Transferable but value depends on secrecy | Creates the chain of title enabling transfer |
| Risk under EU AI Act obligations | Disclosure may interact with transparency duties | Record-keeping duties must not breach secrecy | Must reflect compliance warranties and logs |
For startups, investor confidence depends on a clean IP position. The central strategies are: ensure every founder, employee and contractor has assigned all relevant IP to the company; implement founder vesting so that departing founders cannot strip the company of value; and maintain a comprehensive IP schedule listing patents, applications, trade secrets, models, datasets and licences. In a share purchase agreement or shareholders’ agreement, investors will expect warranties that the company owns or has valid licences to all IP it uses, that there are no outstanding assignments pending, and indemnities covering third-party claims. AI model ownership should be documented specifically, including the provenance of training data and the licence terms of any third-party models or components.
The EU AI Act (Regulation (EU) 2024/1689) is being applied in phases following its entry into force in 2024, with key obligations taking effect across 2025 and 2026. While it does not directly change who owns IP, its compliance architecture reshapes how AI-generated inventions Spain are licensed, documented and diligenced.
The European Commission’s information hub and the regulation text on EUR-Lex set out a risk-based framework. The obligations most relevant to IP and commercial drafting include transparency duties, technical documentation and record-keeping requirements, and heightened obligations for systems classified as high-risk. These requirements mean that providers and deployers must be able to demonstrate how systems were built, trained and tested.
The commercial effect is significant. Licences must now specify who holds and maintains the compliance documentation, who bears the cost of conformity obligations, and how logs and records are shared without compromising trade secrets. Warranties and indemnities increasingly address AI Act compliance alongside traditional IP warranties. The likely practical effect, early indications suggest, is that AI Act compliance becomes a standard condition precedent in licensing and acquisition deals.
Acquirers and investors assessing AI-generated inventions Spain should extend traditional IP diligence to cover AI-specific and regulatory risks. A robust checklist includes:
Clear drafting is the most reliable way to secure ownership of AI outputs in Spain. The following checklist and clause headings offer a practical starting point; all drafts should be reviewed by local counsel before use.
Checklist for securing ownership: identify all human contributors and AI tools used; include present-tense IP assignments in every employment and contractor agreement; address statutory employee-invention compensation; separate ownership of models, weights, training data and outputs; obtain warranties of originality and non-infringement; secure disclosure of open-source and third-party components; and align licensing terms with EU AI Act compliance obligations.
A strong employment clause assigns to the employer, upon creation, all economic IP, including inventions created with AI assistance, arising in the course of employment, confirms the employee’s cooperation in perfecting title, and addresses any statutory compensation for employee inventions under Spanish law.
A contractor clause provides for an express, present assignment of all economic IP in the deliverables and AI-generated outputs, a waiver of moral rights to the extent permissible under Spanish law, an obligation to procure sub-contractor assignments, and warranties that deliverables do not infringe third-party rights.
A model and data licence clause defines separately the rights to the model, the weights, the training data and the outputs; specifies permitted uses and territorial scope; allocates responsibility for AI Act compliance documentation and record-keeping; and includes indemnities for infringement and data protection breaches. Key negotiation points typically include indemnity caps, the scope of output ownership, and which party retains improvements made during the term.
When ownership of AI-generated inventions Spain is contested, several enforcement routes are available. Specialised commercial courts (juzgados de lo mercantil), including the commercial courts in jurisdictions such as Barcelona and Madrid, hear IP and patent disputes, and parties can seek preliminary injunctions and measures to preserve evidence, particularly important for AI assets where models and training records can be altered or deleted. Preservation of evidence measures can secure source code, model weights and logs before they disappear.
Arbitration is a common alternative where contracts provide for it, offering confidentiality that can be valuable for trade-secret-heavy AI disputes. Remedies range from injunctions halting infringing use to damages for losses suffered. A recurring practical challenge in AI disputes is evidence: proving contribution, provenance and misappropriation often requires technical analysis and well-maintained documentation, which underscores the value of contemporaneous record-keeping. Well-drafted contracts that fix ownership and include clear dispute-resolution mechanisms reduce both the likelihood and the cost of litigation.
Securing ownership of AI-generated inventions Spain in 2026 requires a coordinated legal, commercial and compliance strategy. The machine will not own the invention, but the humans and entities around it can, or can lose it, depending on how rights are allocated. Act on five priorities:
For tailored advice on AI-generated inventions Spain, including bespoke clause drafting and transaction support, Global Law Experts can connect you with qualified local counsel. You can also explore the Technology practice area and the Technology lawyers in Spain directory on Global Law Experts for further resources.
This article was produced by Global Law Experts. For specialist advice on this topic, contact Jesus Osuna at Addwill, a member of the Global Law Experts network.
posted 4 minutes ago
posted 24 minutes ago
posted 45 minutes ago
posted 1 hour ago
posted 1 hour ago
posted 1 hour ago
posted 1 hour ago
posted 2 hours ago
posted 2 hours ago
posted 3 hours ago
posted 3 hours ago
posted 4 hours ago
No results available
Find the right Legal Expert for your business
Send welcome message