A Practice Note discussing key issues concerning patents and AI inventions, including machine learning, neural networks, and deep learning technology, and their applications. This Note discusses recent AI patenting trends, potential types of AI-related inventions, and key issues and best practices for obtaining patent protection for AI-related inventions in the United Arab Emirates (UAE). It also discusses AI inventorship, novelty, inventive step, and potential infringement considerations for patents and applications claiming AI-related technology.
AI innovation, including techniques for computation, speech and image recognition, predictive analysis, and their applications in different fields, has become increasingly common and consequential across many technologies. Through vast amounts of data and complex computer algorithms, computers are now able to better recognize objects and make predictions and play an important role in robotics, automated transportation, natural language processing, telecommunication routing, and other areas. For example, AI-related inventions provide tools that promise to improve the efficiency and effectiveness of medical research, and, ultimately, diagnoses and treatments. Despite the critical importance of AI-related inventions to these and other technological advances, current UAE patent law is still developing, including in areas addressing the patentability of AI-related inventions. This uncertainty risks incentivizing potential patentees to keep certain techniques of important AI-related inventions secret rather than seek patent protection, reducing the open exchange of information that is fundamental to the UAE patent system.
This Note discusses key issues and best practices regarding patents and AI-related technology, including:
• Recent AI patenting trends in the UAE.
• Key issues and best practices concerning AI-related technology and patents in the UAE, including:
– patent-eligible subject matter;
– sufficiency, enablement, best mode, and claim clarity;
– AI inventorship;
– novelty and inventive step; and
– potential infringement issues.
AI and Patents in the UAE
The UAE has positioned itself as a regional leader in AI adoption and innovation. In 2017, it appointed the world’s first Minister of State for AI and launched the UAE National AI Strategy 2031, which aims to make the UAE a global hub for AI technologies and legislation. The government and private sector have invested heavily in AI research.
For example, the Mohamed bin Zayed University of AI(MBZUAI) in Abu Dhabi was established as a graduate-level research institution dedicated to AI. As the UAE develops its own AI solutions, observers anticipate a rise in patent applications for AI-related inventions filed in the UAE.
Globally, patent filings in the AI field have grown remarkably, and the trend is likely to be reflected in the UAE’s patent filings as well. Many AI patents worldwide are filed by tech companies (for example, IBM, Microsoft, and Google), and these companies may seek protection in the UAE for important innovations, especially given the UAE’s emphasis on becoming a technology and innovation hub. While specific UAE Patent Office statistics on “AI patents” are not publicly available, unofficial evidence suggests increased interest in protecting AI-driven innovations (for instance, inventions in fintech, smart cities, and healthcare AI applications) under the UAE patent system.
However, the UAE’s legal framework for patents only recently underwent significant reform. Federal Law No. (11) of 2021 on the Regulation and Protection of Industrial Property Rights (UAE Patent Law) came into force in 2022, replacing the previous 2002 patent law. This new law aligned many patent standards with international norms (for example, introducing a grace period for certain types of inventor disclosures).
However, it also explicitly clarified the exclusions from patent-eligible subject matter, notably including a bar on patenting software. Since many AI inventions are software-based, this statutory exclusion presents a key challenge for AI patenting in the UAE. Early indications are that patent applications for AI-related inventions in the UAE must be crafted carefully to meet the new law’s requirements. Currently, there are no reported UAE court decisions specifically dealing with AI-generated inventions or AI inventorship, meaning practitioners must infer likely outcomes from the statute and from international trends. Overall, the trend is that the UAE is encouraging innovation in AI, but patent protection for such innovations requires navigating within the boundaries of UAE law.
AI Subject Matter Eligibility
To qualify for a patent in the UAE, an invention must be:
• New.
• Involve an inventive step.
• Be industrially applicable.
(Article 5, UAE Patent Law.)
These general criteria apply to AI-related inventions as well. In practice, an AI invention (such as a new machine learning method or an AI-enhanced device) must be:
• Novel (that is, not previously disclosed to the public).
• Non-obvious (that is, not an obvious development for a skilled person in view of prior art).
• Useful or industrially applicable (that is, capable of being made or used in some kind of industry).
Importantly, the UAE has a 12-month grace period for inventor disclosures (meaning an inventor’s public disclosure up to 12 months before the filing date will not count against novelty). This grace period can be particularly relevant for AI innovations, where researchers often publish papers. However, patent applicants relying on this grace period should be cautious if international filings are planned, since not all countries offer a grace period. To view and customize comparison charts on patent application filing grace periods, see Quick Compare Chart, Patent Prosecution (available to PL Dynamic subscribers).
Excluded Subject Matter (Software and Algorithms)
A critical issue for AI patents is the scope of patent-eligible subject matter, given that AI inventions often involve software or mathematical techniques. UAE law explicitly excludes certain subject matter from patent protection like software and algorithms. Article 7(1) of the UAE Patent Law lists categories of inventions that “shall not be granted” patents or utility certificates, including:
• Scientific discoveries and theories.
• Mathematical methods.
• Plans, rules, and methods for performing mental acts or conducting business.
• Computer programs (software).
In other words, software “as such” is not patentable in the UAE. This is a significant provision for AI, since AI platforms and algorithms are typically implemented in software. Absent a technical embodiment, a pure AI algorithm or trained model would likely fall under the excluded categories of a mathematical method or computer program. This aligns with the approach in many jurisdictions (for example, the European Patent Convention also excludes mathematical methods and software per se from patentability).
For information on patent subject matter eligibility for software in Europe, see Practice Note, Patents for Computer-Implemented Inventions in Europe.
The software exclusion does not mean that all AI-related innovations are not eligible for patent protection. Rather, it means that the patent claims must be framed in a way that the invention is not merely an abstract algorithm. The UAE Patent Office will look for either:
• A technical aspect.
• An applied invention.
For instance, an AI-based technique that improves a concrete technical process or is embedded in a physical device may be patent-eligible, whereas an attempt to patent a standalone algorithm or a general AI concept will be rejected. The law itself does not elaborate on any “technical effect” exception to the software rule, but in practice UAE examiners are expected to follow international practice for computer-implemented inventions. In other patent systems (for example, US and Europe), a common strategy is to emphasize how the AI invention produces a technical improvement, for example:
• Reducing network latency.
• Improving image processing accuracy.
• Efficiently using computing resources.
For information on AI patent eligibility in the US, see Practice Note, AI: Patentability Considerations.
In the absence of local case law or examination guidelines on AI, applicants should proactively provide explanations in the patent application (or in response to any office action) as to how the AI-related invention produces a technical benefit. Citing examples from EPO or USPTO prosecution where similar claims were allowed can be effective. So far, UAE examiners have been receptive to well-founded arguments that an AI invention is more than a mathematical method, especially if one can point to concrete embodiments or experimental results showing a technical improvement.
It is also important to remember formal requirements. The UAE requires patent specifications to be filed in Arabic and English. Ensuring that complex AI terminology is accurately translated into Arabic is crucial. Ambiguities in translation could potentially lead to misunderstandings about the invention’s scope (for example, mistranslating a term like “neural network” or “classifier” could inadvertently broaden or narrow the interpreted subject matter). Patent attorneys in the UAE often work closely with specialized translators to preserve the technical meaning in the Arabic filing. This is a practical aspect of prosecution that can significantly affect how an examiner perceives an AI invention.
In summary, while no UAE-specific examination manual for AI exists yet (and there are no reported UAE court decisions specifically dealing with AI-generated inventions), applicants should assume a careful scrutiny of AI patent applications. By aligning the application with international norms and clearly conveying the practical, technical nature of the invention (in both language versions of the filing), applicants can navigate the UAE Patent Office process more effectively.
AI Subject Matter Eligibility Best Practices
To maximize the chances of an AI-related invention being eligible for patent protection in the UAE, practitioners should consider the following drafting strategies:
• Emphasize a technical application or problem-solving aspect. Frame the invention as a solution to a technical problem in a specific field (for example, “a machine learning method for detecting manufacturing defects in a production line” rather than just “a machine learning algorithm for pattern recognition”). Highlight any improvement in performance, efficiency, or functionality in a tangible process or machine.
• Avoid claims that only generally references the AI algorithm. Instead of claiming “a neural network that learns to classify X,” it may be better to claim “a system comprising [hardware] configured to perform [specific functions]” or “a computerimplemented method for [task], comprising [steps]” with those steps tied to real-world inputs or outputs. This helps ensure the invention is seen as an applied technological development, not a disembodied mathematical method.
• Include any necessary hardware or physical components. If the AI runs on a specialized device or uses particular sensors or processors, include those in the claims. Even if the core novelty lies in the algorithm, anchoring it to a device can help overcome the “pure software” objection. For example, claim a “control system for a robotic arm using an AI algorithm to optimize movement” rather than the algorithm by itself.
• Leverage guidance from other jurisdictions. In the absence of formal local guidance, consider the EPO’s two-hurdle test (that is, require a technical character and a technical contribution) or the approach under US law (that is, avoid claims that could be considered to be “abstract ideas” by ensuring the claims contain inventive concepts). While UAE law is distinct, such approaches inform what an examiner might find acceptable. In practice, UAE examiners often rely on international search and examination reports (for PCT national phase entries) and the standards applied therein.
By carefully drafting AI patent applications in this manner, applicants can improve the likelihood that the invention will be viewed as patent-eligible subject matter under UAE law, despite the explicit exclusion of software and algorithms. It is also advisable to stay informed of any UAE Patent Office practice updates or guidelines on AI inventions, as this is an evolving area.
Patent Disclosure Requirements
Patents are granted in exchange for public disclosure of the invention. This means an AI-related patent application in the UAE must meet the same disclosure standards as any other invention to ensure that a person skilled in the art could carry out the invention based on the patent. UAE law explicitly spells out key disclosure requirements:
• Sufficiency of description (enablement) (see Sufficiency).
• Whether the disclosure enables implementation of the claimed invention across the full scope of the claims (see Enablement).
• Clarity and support of claims (see Claim Clarity).
UAE patent applications are examined for sufficiency and clarity during prosecution, and lack of adequate disclosure can lead to an office action rejecting claims in the application. It is generally easier to provide a comprehensive disclosure upfront rather than to try to add information later since adding new matter after filing the application is not allowed.
UAE patents cannot be filed through the GCC Patent Office. Applicants seeking patent protection in the UAE must file directly with the UAE Patent Office under the national procedure. There is no current indication that the UAE will allow filings through the GCC Patent Office in the near future. For information concerning patent application filing and prosecution through the GCC, see Practice Note, Patent Application Filing and Prosecution through the Gulf Cooperation Council.
While UAE law does not use the term “best mode” explicitly as in US law, UAE law does require that the patentee disclose the “best way possible” to carry out, and fully enable, the claimed invention (Article 11(5), UAE Patent Law). This requirement implies that if the inventor knows of a preferred way to implement the claimed invention, it should be included in the description. For example, if a certain neural network configuration yields the best results, or a particular training hyperparameter is critical, these details should be disclosed.
Claim Clarity
Claims must be clear, concise, and fully supported by the description (Article 11(7), UAE Patent Law). This requirement (similar to claim definiteness and support) is crucial in AI patents because terminology can be obscure. Every claim element (such as “trained model,” “classification engine,” or “neural network”) should be explained in the description. Lack of clarity (for instance, a claim referring to an “intelligent system” without definition) can lead to rejection or later invalidation for indefiniteness.
For information on claim drafting for computer-implemented inventions in the US, see Practice Note, Patent Drafting: Claiming for Proper Interpretation in Computer-Implemented Inventions.
AI Patent Disclosure and Claim Drafting Best Practices
AI terminology should be used consistently and preferably defined in the description. If the claims mention terms like “machine learning model,” “training module,” or “inference engine,” the description should clearly explain what those mean in the context of the invention.
Additionally, an applicant should avoid purely functional claiming that might render the scope indefinite. For example, a claim reciting “a device that uses AI to achieve optimal results” is vague. In this case, the patent specification should identify the type of AI and the metric defining “optimal.”
Finally, the use of established terminology (such as referring to standard AI techniques) can be helpful in a patent application. However, an applicant should ensure that any non-Arabic terms match what would be reflected in the UAE translation and avoid the potential trap of simply translating technical terms into Arabic. An applicant should verify that the chosen Arabic term precisely conveys the intended meaning to avoid later ambiguity.
Inventorship
In the UAE, a patent can only be granted to an inventor or someone who derives rights from an inventor (such as an employer or assignee). The inventor’s name must be stated in the application (Article 8(1), UAE Patent Law). If multiple people contributed, they may be joint inventors, and someone who merely assisted without contributing an inventive step is not considered an inventor (Article 8(4), UAE Patent Law). The concept of inventorship thus relies on making a creative contribution to the invention’s conception.
The UAE Patent Law does not explicitly define “inventor” as a natural person, but all language in the statute assumes a human inventor (for example, using pronouns like “he” and discussing rights of the inventor and their successors). In practice, patent offices and courts worldwide have so far interpreted “inventor” to mean a human person, and the UAE is expected to be no different.
Until and unless laws are revised to acknowledge non-human inventors, all AI-related patents in the UAE must have human inventors. Patent professionals should take care to identify the proper inventors and document their inventive contributions, as errors in inventorship can jeopardize patent validity. They should also watch global developments, as the debate on AI inventorship is ongoing. Although the consensus is currently that AI cannot be an inventor, any future legal changes (in the UAE or abroad) would significantly impact how patents for AI-generated inventions are handled in the UAE.
AI Inventorship
The question of whether AI can be an inventor has been tested in other jurisdictions through the famous DABUS case, where an AI system was named as the inventor on patent applications. Every major patent authority to date (for example, US, UK, and the EPO) has rejected the notion of a non-human inventor, often on the basis that patent laws were intended to reward human ingenuity. The fundamental rationale is that patent systems are designed to incentivize people. An AI, lacking legal personality and the ability to transfer rights except via its owner, does not fit into the scheme of inventorship and ownership.
To view and customize comparison charts on AI inventorship, see Quick Compare Chart, AI and Patents (available to PL Dynamic subscribers).
The UAE has not yet faced a published case of an AI being named as an inventor. Given the statutory framework, an application listing an AI (with no human) as the inventor would almost certainly be rejected on formal grounds when the details are verified. The UAE Patent Office would likely require a real person’s name, as it routinely does, and if none is provided, the application cannot proceed to grant.
Thus, AI cannot be an inventor under current UAE law. An inventor must be a natural person who conceived the invention. Even if an AI algorithm autonomously produced a novel solution, an applicant cannot legally list the AI itself on the patent. This creates a practical issue of how to handle inventions that are significantly generated by AI. The recommended approach is to identify the human who qualifies as the inventor in such scenarios. This is typically the person (or team of people) who planned, programmed, or recognized the significance of the AI’s output. If, for example, a researcher set up an AI system to design new materials and the AI outputs a new material composition, the researcher who set the process in motion and selected the result could be seen as the inventor (since they had the “innovative idea” to solve a problem using that system). It may feel like a legal fiction in cases of near-total AI autonomy, but until laws change, naming a human inventor is mandatory to obtain a valid patent.
Inventors in the UAE have the right to be acknowledged and to assign their rights. If no inventor is named (or if a non-human is named), the patent is not enforceable. Therefore, patent attorneys should counsel their clients to:
• Always name at least one human inventor.
• Ensure that person’s contribution meets the threshold of an inventive contribution (not just a minor involvement).
• If multiple team members worked with the AI, carefully evaluate who contributed to the conception of the claimed inventive concept.
Employee Inventions and AI
In many AI development scenarios, the “inventors” are employees of a company (for example, data scientists or engineers developing AI solutions). UAE Patent Law addresses ownership of employee inventions in Article 10.
Generally, if an invention is made in the execution of an employment contract (and the contract does not say otherwise), the employer is the default owner of the patent rights in the UAE. The inventor-employee still is named as the inventor, but the patent can be filed by or assigned to the employer. Additionally, if an employee not hired to invent nevertheless creates an AI invention using the employer’s resources, the employer can claim ownership by notifying the employee of its intent within a certain period, with the employee entitled to fair compensation. For information on employee-inventor remuneration, see Practice Note, Inventor Remuneration (UAE).
In practice, companies developing AI in the UAE should have clear invention assignment agreements with their developers, consistent with the law, to ensure smooth ownership of AI-related patents. From an inventorship perspective, it is wise to document the contributions of team members during AI research and development projects. For example, companies should keep records of who designed the algorithm, who tuned it, and who interpreted the results. This can help later in determining who should be listed as an inventor on a patent application and avoid disputes.
Novelty and Inventive Step
AI inventions must clear the normal hurdles of novelty and inventive step for patentability. Practitioners should thoroughly document and claim what is technically new and non-obvious about their AI invention, whether it is a new algorithm, an unexpected application, or a superior performance that required creative insight.
Given the fast pace of AI developments, ensuring novelty can be challenging, so timely filing and comprehensive prior art searches are essential. For inventive step, applicants should focus on concrete technical contributions, not just the idea of using AI, to demonstrate that the invention is a leap beyond the prior art and not merely an obvious next step.
Novelty
An invention is novel if it has not been publicly disclosed before the effective filing date (Article 5(3), UAE Patent Law). Under the UAE’s 12-month grace period, disclosures of information by the inventor, or another party who obtained the information, directly or indirectly, from the inventor, is not prior art against a subsequently filed patent application if filed within 12 months after the inventor disclosure (Article 5(4), UAE Patent Law). For information on the filing grace period in the Gulf Cooperation Council Patent Office, see Practice Note, Patent Application Filing and Prosecution through the Gulf Cooperation Council: Filing Grace Period.
For AI inventions, sources of prior art may include:
• Academic research papers.
• Open-source code repositories.
• Conference presentations.
• Patents.
• Products.
AI as a field is characterized by rapid publication and sharing of results (for example, on preprint servers or at international conferences). This means the bar for novelty can be quickly compromised if inventors are not careful. Note the common scenario where a research team develops a new AI technique and publishes a paper about it, then later decides to file a patent application. In the UAE, if that publication was by the inventors (or someone who obtained the information from them) within the 12 months immediately preceding the patent application’s effective filing date, the grace period provision negates any prior art affect the publication would have otherwise created, maintaining novelty for the new AI technique.
However, if the publication was earlier than 12 months or by an unrelated party not obtaining the information from the inventor, the publication will be prior art and the invention is not novel. Best practice is to file patent applications before any public disclosure whenever possible, even though the grace period is available as a safety net.
Patent applicants should keep in mind that if they intend to file outside the UAE, some jurisdictions may not honor the grace period. Therefore, an early filing strategy is wise for global protection. For information on global patent filing strategies, see Practice Note, Global Patent Prosecution: Filing Considerations.
Another aspect of novelty is that AI inventions might be incremental improvements or combinations of existing techniques. For instance, using a known machine learning algorithm on a new type of data might or might not be novel, depending on whether someone has done that before. It is crucial to perform a thorough prior art search to ensure the invention has not already been described, including searching:
• Patent databases.
• AI literature.
• Online stores (for example, arXiv papers or GitHub projects).
A potential obstacle is that AI implementations are sometimes disclosed informally (for example, as open-source code) without fanfare, but such disclosures can still be prior art if accessible to the public.
Inventive Step
An AI invention involves an inventive step if it is not obvious to a person skilled in the art in light of the prior art (Article 5(5), UAE Patent Law). The law frames this as not being an “intuitive procedure” for an ordinarily skilled person in the relevant field. For AI-related patents, the PHOSITA could be considered an AI engineer or a domain expert using AI, and the prior art includes existing algorithms and their known applications (see Level of Ordinary Skill in the Art).
Key considerations for inventive step in AI inventions include:
• Combining known AI techniques. Many AI inventions combine known components (for example, models or algorithms) in a new way or for a new purpose. Examiners will ask if it would have been obvious to do this. For example, if someone applies a well-known image recognition algorithm to solve a known problem in agriculture (like identifying crop diseases), is that a non-obvious invention or just an obvious application of known technology? The answer may depend on whether there were unexpected difficulties or whether prior art taught away from using AI in that context. If the use of AI yields surprising results or requires non-trivial adaptation, it can be argued as nonobvious. Applicants should highlight in the patent what technical hurdles had to be overcome or why the solution was not straightforward.
• Parameter tuning and training improvements. AI innovation often lies in tweaking model architectures or training methods. However, simply adjusting parameters (for example, making a neural network deeper or using more training data) might be seen as routine experimentation, hence obvious. To establish inventive step, applicants should emphasize any specific insight or advantage, such as a novel training algorithm that achieves results that conventional training could not, or a new model structure that is counter-intuitive yet effective. Quantitative improvements can support an inventive step argument, but an applicant should still identify a technical reason for the improvement in the disclosure.
• Automating tasks previously done by humans. Many AI inventions involve automating something via AI. For inventive step, if the task itself is known and the idea of automating it using a computer is obvious, then the mere use of AI might be considered obvious absent more detail. Patents have been denied in some jurisdictions for claims that essentially say “do this mental task using AI.” To be inventive, the implementation should have some complexity or unique adaptation. For instance, using AI to diagnose medical images could be obvious in concept (doctors already diagnose images mentally, so automating that is a logical goal), but a specific AI solution that achieves it can be inventive if it was not straightforward to find. It is the difference between a broad goal and the concrete means to achieve it.
• The role of data. AI results can depend heavily on training data. However, providing a new dataset to a known algorithm is usually not a patentable inventive step (datasets themselves might not be patentable unless they result in a specific improved system). If an invention relies on a special dataset, an applicant should consider if the invention can be characterized as the trained model or the training method rather than the data itself. The inventive step could lie in recognizing which data features are important for the model, or in a feature engineering method.
One emerging consideration is whether the “person skilled in the art” should be assumed to have AI tools at their disposal (in other words, would a skilled person be expected to use AI to solve problems, thus potentially rendering some AI-assisted solutions obvious?). Currently, the standard skilled person is a hypothetical human technician, not an AI. But for inventive step, examiners might argue that if a problem could be solved by routine use of machine learning by a skilled practitioner, the outcome might be obvious. For instance, optimizing a known product design using generic AI optimization might be seen as something any skilled engineer could do given modern tools. To counter such arguments, an applicant should identify any non-routine aspects, for example:
• Modification of the AI tool in a non-obvious way.
• The AI produced a result that experts would not have expected.
Emphasizing those points in the patent specification can be crucial for establishing an inventive step.
In UAE practice, the inventive step analysis is similar to European practice in many respects. There is no formal “problem-solution” test codified in law, but framing the inventive concept in terms of a problem solved can help. An applicant should:
• Identify the closest prior art, explain the problem or limitation with it, and show how the invention provides a solution that was not implied by prior art.
• Clearly indicate if the invention yields an unexpected technical effect. Unexpected results are strong evidence of non-obviousness.
For information on inventive step in Europe, see Practice Note, Patentability: Inventive Step.
AI Patent Infringement
Once an AI patent is granted, the owner has powerful rights to prevent others from using that patented AI technology in the UAE market. Enforcing those rights might require technical detective work due to the intangible nature of software. The UAE Patent Law provides the legal tools for enforcement (see Patent Infringement), and there are no AI-specific patent infringement exemptions (see Infringement Exceptions).
Patent holders should be prepared to explain their technology to courts and possibly educate experts. Alleged infringers should be ready to show that their technology is different (non-infringing) or to challenge the patent’s validity (for example, by arguing that the patent claims are too broad or anticipated by prior art, both of which are common defenses in infringement cases).
For information on patent infringement litigation generally in the UAE, see Practice Note, Patent Litigation: Overview (UAE).
Patent Infringement
The UAE Patent Law provides patent holders with the exclusive right to exploit the invention and to prevent others from exploiting it without consent (Article 19, UAE Patent Law). This includes, for a patented product, the right to stop others from making, using, selling, or importing that product without permission. For a patented process (for example, an AI-implemented process), the law gives the patent owner rights over use of that process and over the products directly obtained by that process.
In essence, any commercial or industrial use of a patented AI invention by an unauthorized party constitutes infringement. The patent holder can pursue legal action to stop the infringement (typically through civil court injunctions and damages in the UAE). There are also criminal penalties in the UAE for willful patent infringement under certain circumstances, which reflects the seriousness of IP violations.
Infringement in the context of AI can be tricky to detect and prove. AI inventions often reside in software or behind the scenes of a service. For example, if a company secretly uses a patented AI algorithm on its servers to gain a competitive advantage, the patent holder may not easily know this is happening. Unlike a physical product that can be examined for patent-covered features, an AI process might require access to the adversary’s software or system to prove infringement. This is where litigation tools like discovery (in common law jurisdictions) or court-appointed experts (in civil law jurisdictions like the UAE) come into play. In a UAE court, an IP owner can request the appointment of an expert to examine the defendant’s product or code for infringement, though obtaining such an order requires convincing the court that there is a solid initial case.
Another facet is direct versus indirect infringement. UAE law does not explicitly specify contributory infringement, but generally any party that carries out the prohibited acts (that is, manufacture, use, sale, or import) is directly liable. If a third party supplies a component specifically for use in an infringing AI system (such as specialized software modules solely to perform a patented AI method), that third party could potentially be liable under general tort principles or complicity, even though the law does not have a specific contributory infringement provision. Practically, patent holders target the end user or provider of the AI system.
For information on the scope of patent protection and patent infringement, see Practice Note, Scope of Patent Protection (UAE).
Infringement Exceptions
The UAE Patent Law provides exceptions where certain acts do not count as infringement even if they involve the patented invention.
Notably, acts done for scientific research or educational purposes are excluded from the patent holder’s rights (Article 22(1), UAE Patent Law). This means if a university or a researcher in the UAE is using a patented AI algorithm purely for academic research (and not for commercial exploitation), that use would not be considered infringement. This exception is important in the AI world because AI research often builds on existing algorithms. Researchers can study and experiment on a patented AI technique without permission, if it is non-commercial and for knowledge advancement. However, the moment it crosses into commercial deployment or product development, the exception no longer applies.
Another exception in Article 22(2) of the UAE Patent Law relates to use of patented inventions on vessels or aircraft that temporarily enter the country (a kind of international transport exception), which is less directly relevant to AI unless software is being used on a plane in transit.
UAE law also recognizes a form of prior use right (Article 20, UAE Patent Law). If someone had already in good faith been using an invention in the UAE or made serious preparations to do so before the filing date of the patent by another, they can continue that prior use. For AI, this could apply if a company was already using a particular AI technique internally and then another party patents it later. The first company might have the right to keep using it (though they cannot expand beyond the scope of the prior use). This is a narrowly applied principle requiring evidence of the earlier use.
Enforcement Process
Patents in the UAE are enforced through the civil courts (IP enforcement is under federal jurisdiction). The patent must be registered and in force. Typically, a patent holder sends a cease-and-desist letter first. If that fails, a lawsuit for infringement can be filed. The court may appoint an independent technical expert to evaluate infringement and validity issues and report back. Remedies include injunctions (to stop the infringement) and damages (compensation for losses).
In some cases, customs authorities can be engaged to seize infringing goods at borders if it is a product patent. However, this is less applicable for AI methods unless embodied in exported software or devices.
Special Issues with Enforcement of AI Inventions
If an AI system itself “decides” actions that lead to infringement, the user of the system is still the liable party. For instance, if someone deployed an AI that inadvertently replicates a patented invention (such as where an AI that designs solutions ends up designing something covered by another’s patent), the person running the AI would be responsible for any resultant infringement since there is no concept of the AI being the infringer on its own. This again ties back to the view in the UAE that humans control and use AI.
Freedom-to-Operate and Risk Management
Companies implementing AI in the UAE should conduct freedom-to-operate (FTO) analyses to ensure they are not unknowingly infringing existing patents. Traditionally, only granted patents were published in the UAE. However, under the new practice, the UAE Patent Office now publishes patent applications 18 months from the effective filing date. FTO reviews should therefore combine the following:
• Searches of UAE-published applications and granted patents.
• Monitoring of international filings (including PCT applications and foreign national filings) that may enter the UAE national phase.
• Ongoing surveillance of competitors’ portfolios.
Because this 18-month publication is relatively recent, some pending applications may still be unpublished. Monitoring foreign filings remains a useful early warning tool.
For more information, see Abu Dhabi Intellectual Property Unit: What is a Patent: Patents’ Lifecycle and Timelines.
Other IP Protection for AI Inventions
Because of the limitations in patent protection for certain AI-related subject matter (notably, the exclusion of software and algorithms), inventors should also consider alternative forms of IP protection. In the UAE, the main alternatives are:
• Trade secrets (see Trade Secrets).
• Copyright (see Copyright).
Each offers protection for different aspects of AI technology and can complement a patent strategy or substitute for patents when an invention is not patent-eligible. If an AI innovation is not patent-eligible (for example, it is a pure algorithm or an improvement in code efficiency), an inventor could rely on trade secret and copyright. If it is patenteligible, but an inventor decides the disclosure would too greatly aid competitors or the invention could be kept internal (for example, a manufacturing AI process used in-house), an inventor might opt for trade secret protection instead of patent protection.
Patents are most valuable when an inventor must disclose an innovation to enable a product but wants to prevent others from implementing the same concept even if they independently develop the same technology. Trade secrets are valuable when the goal is to maintain exclusivity by silence, and the risk of independent development is low or acceptable.
Copyright is somewhat automatic, as any AI code an inventor writes is protected. But an inventor might choose to enforce copyright in scenarios where patent protection is not available or is narrower. For instance, if a competitor does not literally infringe the patent, such as where the competitor could design around the claims but in doing so they copied substantial portions of the original code, an inventor could pursue a copyright infringement claim.
A recommended strategy for AI innovators in the UAE is to use a combined approach of IP protections by:
• Patenting the aspects of the AI invention that qualify for patent protection, especially if they involve a new technical application or device.
• Maintaining proprietary know-how as trade secrets, which require strong internal confidential policies as required by law.
• Leveraging copyright to protect software implementations from outright copying.
This multi-layered approach ensures that whether or not a particular AI development is patentable, the innovation can still be safeguarded to some extent. Moreover, having overlapping IP rights can be advantageous. For example, if someone infringes a patent and also copies the code, an inventor could assert both patent and copyright infringement claims in a single legal action.
Patent attorneys should counsel clients on alternative IP protections, especially when dealing with inventions where patent eligibility may be a challenge, such as where the invention comprises a new algorithm. Sometimes the best “patent” strategy might be not filing a patent at all but using secrecy. One could theoretically keep an invention secret during development and only obtain a patent later if needed, leveraging the grace period if the inventor accidentally published information about the invention. Each approach has risks and benefits, so decisions should be made on a case-by-case basis considering the nature of the AI innovation, its expected lifespan, and the competitive landscape.
Trade Secrets
A trade secret is essentially confidential information that gives its owner a competitive edge. Many AI algorithms, models, and especially training data or processes can be kept secret rather than being disclosed in a patent. UAE law explicitly protects undisclosed information (trade secrets) under the UAE Patent Law. To qualify for protection, the information must:
• Be secret (not generally known to persons in the relevant field).
• Have commercial value due to its secrecy.
• Be subject to reasonable efforts to keep it secret. (Article 61, UAE Patent Law.)
If these conditions are met, unauthorized acquisition, use, or disclosure of the secret by others is unlawful and can be subject to a claim of trade secret misappropriation in a legal action in court. In fact, UAE law considers certain acts like industrial espionage or breach of confidentiality agreements as unfair competition offenses in relation to trade secrets, and the Federal Law by Decree No. (31) of 2021 Promulgating the Crimes and Penalties Law (UAE Penal Code) imposes penalties for disclosing trade secrets in certain cases.
UAE law treats unauthorized disclosure of trade secrets as a serious offense, both under criminal law and other statutes. Anyone who, due to their profession, craft, status, or role, is the custodian of a secret and divulges it without permission faces imprisonment of at least one year, a fine of at least AED20,000 (approximately USD5,400), or both (Article 432, UAE Penal Code). In cases where the offender is a public official who learned the secret in the course of duty (or a person in charge of a public service), the penalty is more severe and up to five years of imprisonment can be imposed (Article 432, UAE Penal Code). These criminal penalties apply whether the person disclosed the secret for personal gain or for another’s benefit, so long as the disclosure was unauthorized.
In addition to the UAE Penal Code, the following UAE laws also stipulate penalties for trade secret misappropriation, reflecting its status as an unfair competition or IP violation:
• The Federal Law No. (11) of 2021 on the Regulation and Protection of Industrial Property Rights, which governs patents, designs, and undisclosed information, sets hefty fines for trade secret violations. Courts can impose fines starting from AED100,000 (approximately USD27,000) for unlawful use or disclosure of a trade secret under this law. This is part of treating trade secret theft as a form of unfair competition or infringement of IP rights.
• The Federal Decree Law No. (32) of 2021 on Commercial Companies (Companies Law) also contemplates protection of business secrets in certain contexts (for example, directors or partners misusing company confidential information). Under the Companies Law, the minimum fine for divulging a company’s secret is AED50,000.
For AI developers, trade secrets can protect items such as:
• Source code of AI software.
• Algorithms.
• Model parameters.
• Proprietary datasets or data-preprocessing methods.
Unlike patents, trade secrets do not require registration and have no fixed term. Protection lasts as long as the information remains secret. This is advantageous for aspects of AI that might not meet the criteria for patentability or that one prefers not to publish. For instance, if a company has a unique algorithm that is difficult to reverse engineer, keeping it secret avoids the disclosure that is required to obtain patent protection, and the company can maintain exclusive use indefinitely (or until someone else independently develops something similar).
However, a trade secret does not prevent independent development by others. If a competitor lawfully develops the same AI technique on their own, another party’s trade secret covering that technique does not prevent the competitor from practicing that trade secret, whereas a patent could.
Trade secrets also can be lost through leaks or if employees subsequently work for a competitor without robust non-disclosure agreements (NDAs). The UAE supports employee mobility, and an employee is free to work for their employer’s competitor in general, unless a well-drafted, reasonable non-compete agreement is in place and applicable. Such a non-compete agreement should be limited in time to no more than two years, restricted in geography, and be specific as to the scope of the non-compete agreement.
Therefore, robust internal controls are needed, for example:
• Confidentiality agreements.
• Security measures.
• Access on a need-to-know basis
UAE law places responsibility on the holder of the trade secret to take reasonable steps to maintain its secrecy. Failing that, the holder of the secret might not be able to claim something was a protectable secret. Essentially, the trade secret holder must behave as one would expect a prudent business to behave when something is valuable and confidential. UAE courts look for concrete steps taken by the trade secret holder to preserve confidentiality.
If the trade secret holder fails to implement obvious safeguards, the courts may refuse to categorize the information as a trade secret. Examples of reasonable measures to maintain secrecy include:
• Using NDAs with anyone given access to the information.
• Marking documents as “confidential.”
• Limiting internal access to the information on a need-to-know basis (perhaps via password protection or other access controls).
• Having internal policies or training that underscore the information’s trade secret status.
• Having physical security (for example, keeping sensitive documents in locked cabinets) and cybersecurity for digital data.
Hybrid Patent and Trade Secret Strategy
In practice, many companies use the hybrid IP strategy of patenting what is patentable (and what can be reverse engineered easily from a product) and keeping other aspects as trade secrets. For example, a company might patent the overall system and usecase of an AI invention (thereby disclosing it broadly) but keep the fine-tuned algorithm parameters or a specific training technique as a trade secret, since that detail might not be apparent from the product itself and does not have to be disclosed.
However, patenting one part of an AI invention and maintaining another part as a trade secret is a delicate strategy. It can be viable, but only if the patented portion is fully enabling on its own. The patent must disclose enough information so that a skilled practitioner could implement the claimed system. The trade secret part should be an ancillary detail or optimization not needed to practice the core invention as claimed. For instance, a patentee might patent the overall architecture or workflow of an AI system and keep a specific tuning of hyperparameters secret. However, the patentee must ensure the patent provides a general way to train the model (perhaps using conventional training methods) so that it works, even if not at the optimal efficiency. If the claimed invention cannot be practiced at all without the secret detail, then the patent would be vulnerable to an enablement challenge (or a sufficiency rejection during examination).
When employing this dual approach, patentees often draft patent claims and specifications carefully so that the trade secret details are not claimed as novel features and are not necessary to meet the claims. They might describe the invention in broader functional terms and indicate that various methods (not necessarily the proprietary one) can be used. This way, they try to avoid an enablement issue. Still, one must be cautious, as overly broad claims without commensurate disclosure can be invalid if the only way to implement the full scope would require undue experimentation or the very secret that was withheld.
Overall, the strategy of partially disclosing an AI invention in a patent and keeping some specifics secret carries a trade-off. It is legal to do (patents set a minimum disclosure, not necessarily every implementation detail), but the patent’s validity will be jeopardized if the withheld material is actually needed to practice the invention or was the best way to do so. Patent counsel should therefore ensure that any trade-secret aspect is truly optional or an improvement not required by the patent claims. If there is any doubt, it is safer to disclose enough in the patent to satisfy the enablement requirement, even if that means revealing more information publicly. Patents can be invalidated for lack of enablement or insufficient description, so patentees must strike the right balance when splitting information between patents and trade secrets.
Copyright
Copyright law protects original works of authorship, which in the context of AI primarily means software code and possibly databases. In the UAE, Federal Decree-Law No. (38) of 2021 on Copyright and Neighboring Rights (Copyright Law) explicitly lists computer programs and software applications as protected work. This means that the source code (and object code) of an AI program is automatically protected by copyright from the moment of its creation, provided it is original. The threshold for originality is quite low (basically, it must be the result of the author’s intellectual effort and not copied). Copyright in software gives the owner the right to prevent others from copying, distributing, or modifying the code without permission. It is a very useful protection for the literal expression of an algorithm.
Software copyright in the UAE lasts for the creator’s life, plus 50 years after death. If the software is owned by a company (work-for-hire), the approximate term is 50 years from publication (unless authorship is later claimed by an individual). During this term, the copyright owner has certain exclusive rights, such as to reproduce, adapt, and distribute the copyrighted software code. After the term expires, the work falls into the public domain.
For information on copyright, see Practice Note, Copyright Litigation: Overview (UAE).
However, copyright only protects the expression, not the underlying ideas or functionality. Two developers can independently write different code that achieves the same AI functionality, and each would own copyright in their code, and neither could stop the other’s program via copyright if no code copying occurred. For AI inventions, this means copyright is a tool to stop wholesale code theft, such as where someone hacks into a server and steals the AI’s source code or a former employee takes code to a competitor, as these would be copyright infringement if they reproduce the code. However, copyright will not prevent a competitor from writing their own code to implement a similar algorithm, if they do not copy the original code.
In addition, copyright protection may also be available for training data and AI outputs. Large datasets could potentially be protected as compilations if they involve creative selection or arrangement (though many datasets might lack that creativity and be more factual). For AI-generated outputs (such as an artwork or a piece of writing produced by AI), the Copyright Law’s broad definition of “work” might be interpreted to include AI-generated content if it meets an innovation criterion, but typically copyright requires a human author. This issue has not yet been decided in the UAE.
Copyright is a supplementary protection for the software implementing an AI invention. Best practice is to ensure that all AI software developed is properly documented and that copyright is asserted (protection is automatic, but companies often include copyright notices and can optionally register important software with the authorities for evidence). In addition, the Copyright Law clarifies that for employee-created works (including software), the employer is the owner by default unless agreed otherwise. This is similar to patents where the employer owns employee inventions. In copyright, this means a company owns the code if it is written by its employees.
Other Forms of IP
Other forms of IP protection include:
• Database rights, although there is no specific database-rights regime in the UAE beyond copyright for databases.
• Contractual protections.
Often, licensing agreements, NDAs, and other contracts are used to protect AI IP when sharing it with partners or clients. For example, when licensing AI software to a customer, a contract to restrict the customer’s use of the software is recommended, which for example should prevent reverse engineering. Design rights and trademarks are generally not applicable to AI algorithms, though user interface designs could be protected via design patents or copyright, and brand names of AI products via trademarks. Those forms of IP are tangential to the core AI technology.
Practical Considerations
Patent practitioners and innovators in the UAE dealing with AI should keep in mind the following best practices to navigate the current legal landscape effectively:
• Thoroughly assess patent eligibility. Before filing, evaluate if the AI invention might fall under excluded subject matter (that is, pure algorithms or software per se). If so, modify the invention disclosure or claim strategy to emphasize technical implementation or specific applied use. Engage in brainstorming with inventors to identify different ways to frame the invention as a technical solution.
• File early and use the grace period cautiously. Given the fast-moving nature of AI research, encourage inventors to file patent applications before publishing papers or releasing code. While UAE law offers a 12-month grace period for inventor disclosures, relying on it can complicate foreign filings and carries the risk of intervening independent disclosures by others. Filing early secures a priority date, which should help in situations where competitors are racing to patent similar AI developments.
• Draft robust and clear patent applications. Ensure the specification provides an enabling disclosure of the AI invention, including examples, diagrams, and detailed descriptions of how the AI works to the extent necessary to teach a skilled person how to implement the claimed invention. Define technical terms and avoid ambiguity. This not only satisfies legal requirements but can also strengthen the resulting patent against prior art invalidity challenges by clearly delineating the inventive features of the claimed invention. Applicants must file applications in both English and Arabic and confirm the correct translation of technical terms.
• Name appropriate human inventors. Always list human inventor(s) who contributed to the conception of the AI invention. Verify inventorship with the research and development team to include all key contributors and exclude those who did not contribute an inventive idea. Document the inventive contributions in lab notebooks or invention disclosure forms in case inventorship is later challenged. Do not list an AI or machine as an inventor. It is not permitted and will render the application defective. When an invention is highly AI-driven, identify the human who guided the AI or recognized the invention.
• Leverage international examination results. For PCT national phase entries or corresponding applications filed elsewhere, use favorable findings from foreign patent offices to support applications filed in the UAE. For example, if the EPO allowed claims on the basis of a technical effect argument, prepare to present a similar argument to the UAE examiner. Conversely, anticipate and preempt issues that arose abroad. For example, if the USPTO rejected claims based on 35 U.S.C. § 101 as only claiming an abstract idea, amend the claims accordingly for the UAE.
• Monitor and adapt to evolving guidance. Stay updated on any new guidelines or practices issued by the UAE Ministry of Economy regarding AI patents. As AI becomes more prevalent, the UAE Patent Office may release examination guidelines or the UAE courts might decide cases that shed light on how to treat AI-related patent issues. Being aware of these developments will help in advising clients correctly. Participate in industry groups or forums on IP and AI in the Middle East to exchange insights.
• Consider hybrid IP protection strategies. Advise clients on using trade secret and copyright protection alongside patents. If a particular innovation is not patentable in the UAE (for example, a core algorithm), ensure the client understands how to protect it as a trade secret (for example, implementing strong confidentiality measures or employee training on data security). Simultaneously, secure copyright in software by simply maintaining records of authorship since formal registration is not required for protection, but can be done for evidentiary purposes. A combination of patent protection for the overall concept and trade secret protection for certain details can sometimes be the best approach for overall IP protection.
• Draft claims with enforcement in mind. When formulating claims, consider how to detect and prove infringement. For AI methods, include system or device claims if possible so that an infringing product can be identified. Avoid only method claims that occur in a black box. If relevant, include claims on the trained model or the trained system’s output. This way, if a competitor’s AI produces a specific tangible output or uses a specific format, that might infringe a claim. This area is complex, but patent counsel should develop claims that can be practically enforced.
• Use examples and analogies in communications. When responding to examiners or explaining issues to clients, use concrete examples to clarify abstract concepts. For instance, analogize an AI algorithm to a new machine part when explaining to an examiner why it has technical character, or use a simple hypothetical to show a client why an AI cannot be an inventor (for example, “If the AI is listed, who signs the oath or declaration? The law expects a human to do so.”). These approaches improve understanding and persuasiveness.
• Prepare for enforcement and licensing. Develop a strategy for enforcement of AI patents. For example, this might involve a protocol for monitoring the market for services similar to a patented AI-based service. Also, consider licensing opportunities as many AI inventions are platform technologies that could be licensed across industries. A patent portfolio can open revenue streams through licensing deals, provided the claims are sufficiently strong and broad. Ensure contracts account for improvements. For example, if licensing an AI technology, the license should clarify ownership of improvements or new versions developed by the licensee.
• Advise on ethical and compliance aspects. While not a direct patent issue, it is important for patent counsel to note that AI is subject to emerging regulatory and ethical guidelines (for example, data protection or bias). Ensure that pursuit of patents, especially in sensitive AI areas (like medical AI), is conducted in compliance with those regimes. A legally sound patent on an AI that cannot be deployed due to regulatory issues would be of limited value. By following these best practices, patent practitioners can better secure and defend patent rights for AI innovations in the UAE. The field of AI and patents will continue to evolve, but a solid grounding in current law combined with strategic foresight will help navigate the challenges. The UAE’s legal environment, though still catching up to some of the novel questions posed by AI, is generally in line with international principles, so many of the strategies applicable in other major jurisdictions are relevant. The key is to adapt those strategies to UAE law and the practical realities of the UAE patent system. Obtaining patent protection for AI-related inventions in the UAE is achievable with careful planning to ensure the invention is characterized in technical terms, disclosed fully, and claimed astutely. At the same time, AI innovators should also consider protecting AI innovations through confidentiality and copyright, given the limits of patentability for software. Patent attorneys in the UAE must play an active role not just in preparing filings, but also in educating inventors about unique issues (such as inventorship and disclosure for AI) and advising on an overall IP strategy. With the UAE’s strong push towards innovation and a knowledge economy, it can be expected that both the volume of AI patent filings and the legal clarity around them will increase in the coming years. Until then, adhering to the principles and practices outlined in this Practice Note will help patent counsel to navigate this cutting-edge intersection of AI and patent law.
Your contacts:
Essa Ziad Galadari (Director)
Khaled Mehdi (Senior Associate)