Impact of Artificial Intelligence on Intellectual Property

Impact of Artificial Intelligence on Intellectual Property

AI and Intellectual Property

Every day, news about Artificial Intelligence (AI) fills the media, often highlighting negative aspects, such as the creation of fake photographs of celebrities or even strikes by screenwriters in Hollywood demanding protection against AI. However, rather than succumbing to the usual pessimism, we will focus here on an overview of how AI is influencing the IP arena and the opportunities it presents.

AI as author

AI has been involved in the generation of inventions for a long time. It has been used internally to inspire ideas, assist in the creation of test scenarios and, most commonly, incorporate AI tools into technical solutions to address specific problems, such as classification, text, image and speech recognition, or big data analytics.

With the recent explosion of generative neural networks such as ChatGPT, Stable Diffusion, DALL-E or Midjourney, AI has reached the general public and these tools have started to be used to generate graphic works and inventions and, although there is no direct record, possibly also industrial designs and device marks. Of course, this has also given rise to questions about copyright infringement.

Clearly, these tools can serve as a starting point or source of inspiration for inventors and authors. However, the current legal framework poses challenges for the direct registration of works or inventions created with AI. Most countries do not recognize “artificial” authorship, and it is still unclear whether individuals will be able to claim authorship in developments obtained directly with these tools. The debate is still open and considerations about the author of the specific commands (prompts) that have guided the generative AI to obtain the result may be decisive.

AI as a patentability/infringement search tool

For some years now, several patent offices have been using artificial intelligence tools in their internal search and classification procedures. For example, the European Patent Office uses them to pre-sort applications and direct them to the appropriate examination units. However, as far as we know, these are basically support tools, while the actual search procedure is carried out by human examination teams with specialized technical profiles.

A simple internet search returns pages and pages of references to AI-powered patent and trademark search tools, although the ecosystem is not yet clear and in many cases the “AI” label seems more of a marketing gimmick or a flashy add-on than a tool with solid foundations and real utility. However, it is more than reasonable to assume that it will not be long before professional AI automated search tools appear that will be able to offer a great speed and breadth of results. These tools would be of great help both in patentability analyses of an invention or trademark registration and in non-infringement studies (freedom to operate).

However, these will almost certainly be private products, black boxes that will return data without the user ever having direct knowledge of what considerations have been taken into account, how the model has been trained or what information is available to the user. More importantly, there is the effect of “hallucinations” due to the fact that deep learning networks are usually designed to always return some result, even if it is not plausible. All of this possibly contributes to unreliable raw results, a multitude of irrelevant references, or failure to find the really important references.

For trademarks, especially for word marks, automated search tools must address idiomatic and cultural considerations, such as phonetics, most representative letters, and interpretation of words in other languages. The quality and accuracy of the search tools will depend largely on how the models are trained and adapted to each of the countries where searches are conducted.

We will have to see how the market evolves, but, a priori, it seems that supervision by an expert will be necessary to filter and review the results, in a process that will surely have to be iterative.

AI as a translation aid

If there is one AI tool that practically everyone has used at some point in time, it is machine translation – are you familiar with Google Translate? Translation engines have been improving a lot over the years and, nowadays, they offer results that are often surprising and allow you to understand the source text quite well. In most cases, even in the field of intellectual property, this type of translation is more than sufficient. For example, when citing a Korean document in a search report, it is common to use machine translation from Korean to English to get an idea of its content. Translations to and from English are often used because most of the major translation engines have traditionally been trained with that language as a base.

Another example: when applying for a European Unitary Patent it is necessary to provide a translation (into English if the language of prosecution has been French or German, or into another EU language if the language of prosecution has been English). Well, in fact, this is a transitory measure, and it is expected to be abandoned in the future when machine translation has improved enough to make such translation unnecessary.

However, there is an important nuance: both translations for the purpose of analysis and those of the Unitary Patent do not have legal effects. When there are such legal effects, translation becomes much more critical and great care must be taken to ensure accuracy. For example, when translating a patent for registration in a particular country, even a small error in the claims can drastically affect the scope of protection or the interpretation of the claims. This problem is even more relevant with machine translations because sometimes the engine returns an apparently syntactically correct sentence that does not really correspond to the meaning of the original text. The translation of texts for legal purposes can be based on an automatic translation, but a linguistic and technical revision is essential to guarantee quality.

AI as a patent drafting aid

AI-assisted drafting is, unsurprisingly, starting to gain momentum with so-called robot patent drafting services aimed at facilitating workflow, helping to structure memory and figures from claims and streamlining tedious tasks. The combination of these tools with so-called large language models (LLMs) promises to reduce the effort required for drafting and provide sufficient quality, provided the process is supervised and reviewed by an expert.

At the moment, these tools are in an early stage, led by startups and very focused on US Patent Office (USPTO) standards. However, AI-assisted drafting is a really interesting field with a future that looks promising for patent agents.


Artificial Intelligence is imbricated in Intellectual Property in a variety of ways, from invention creation to patentability and infringement searches, translation and patent drafting. As we move into the 21st century, it is essential to address the challenges and seize the opportunities that AI presents in this ever-evolving field. Collaboration between AI and human experts will continue to be critical to ensure accurate and high quality results in the IP world.

This article provides an overview of how AI is transforming Intellectual Property and highlights the importance of adapting to this ongoing technological revolution.

Author: Carles Molina Martínez, Telecommunications engineer and european patent attorney.

Note: The conclusion of the article has been elaborated by ChatGPT and the image created with Dall-E.