Artificial inventors: the patent for works created by AI

  1. Home
  2. Artificial intelligence
  3. Synthetic compounds
  4. Artificial inventors: the patent paradox of works created by artificial intelligences

According to a very recent decision fromEuropean Patent Officefollowing the decisions contained thereinUKIPO in England andUSPTO iIn America, a patent cannot be granted to an AI. What are the reasons for this decision? And who can be called an inventor?

Can an AI invent anything?

The answer to this question varies depending on the language used to define the concept. In Italian, for example, the Treccani dictionary defines the inventor as: “the one who invents or has invented something that first creates or finds or even spreads something that did not exist before or was unknown”. According to this definition, anyone who creates something new is therefore an inventor: therefore, even an intelligent machine, if it fulfilled this role, could be defined as such. The speech changes when you try to define the same term in English. Collins’ dictionary, for example, which translates the definition into Italian, says that “an inventor is one person who has invented something, or whose task is to invent things’. Already from the definition we understand that in this case it begins to be complex to frame an artificial intelligence in the context. We are talking explicitly about people who invent, and not about machines that have a similar capability.

It is on these details that we try to update the international regulation in the field of patents, in order to adapt it to a present where not only humans are capable of creating ingenious works, but also machines, if they are specially programmed. The situation is very complex because, on the one hand, there is the market that needs clear rules to attract investment, but on the other hand, there are the responsible bodies that face definitions that are anything but linear. If an algorithm invents something, then who will enjoy the economic benefits? The legal entity that wrote the algorithm? The legal entity that owns the machine that performed it? Who built the hardware? And can AI in itself be considered a legal entity?

AI patents

AI, Invention and Patent: The Invention Machine

Creating autonomous systems capable of inventing something is not a new idea. Back in 1994, Stephen Thaler devised a “Creative Machine”, that is, a neural network that, in interaction with itself and with other machines, was able to generate new patterns of information. Thaler himself has recently invented DABUS (device for autonomous generation of useful information), an AI capable of inventing a food container with a particular fractal shape and an effective light signal for use in emergencies. Thaler, precisely for these inventions, requested the transfer of a patent in the United States, the United Kingdom and Europe, but we will return to this later.

Thaler is not the only one who has tried to develop an autonomous system capable of generating entirely new inventions. John Koza, a professor at Stanford University, for example, started “Invention machine“Which, based on a basic set of information, has invented new electrical circuits, new types of antennas and even lens systems that have never been explored before. The invention machine has a patent, but the ingenuity of this AI can not currently be attributed to anyone directly.

The best practice for modernizing and making the ERP of the future intelligent

Chemistry is another area interested in the revolution of “artificial inventors”. AI and machine learning have become crucial to the pharmaceutical industry as they accelerate the design of new molecules that have the potential to revolutionize modern medicine. This paradigm of scientific discovery represents one of the most exciting frontiers of AI, which in the future may play a key role in research, discovery of new substances for better drugs, materials for more efficient batteries, filtration systems for CO2 reduction, new cells to alternative power supplies and many other things.

AI patent

The legal aspects of the AI ​​invention patent

Some AI pioneers have begun to move to emphasize the importance of regulatory alignment that is able to control the intellectual property of works not created by humans but by machines capable of learn for yourself. Stephen Thaler’s struggle, as we have already mentioned in the previous section, for the grant of the patent for the inventions of his artificial intelligence DABUS, has all the characteristics to enter into the history of technological progress.

Thaler has applied for the patents on the food container and the emergency signal designed by DABUS and has done so in the USA, Great Britain and the EU. In all three cases, the bodies found that the works had all the necessary prerequisites to obtain a patent, but denied the possibility of transferring that patent to an AI.

The USPTO, for example, rejected the application because the rules use specific pronouns to define the owner of the patent, and these pronouns are not compatible with an AI. The conclusion drawn by the Commission is that, according to the case lawinvention is a mental act, and can therefore only be attributed to those who just have a mind, therefore only to a natural person. Thaler also appealed against this decision, but the outcome has not changed.

The EPO’s conclusions are slightly different, although the substance does not change. The European patent authorities did not comment on an AI’s “ability” to create an invention, but simply stated that a machine does not have the legal requirements to be able to claim a right as an inventor. The request was therefore rejected.

In the UK, the form is also different, but the verdict remains the same. UKIPO argued that the inventions can be legally considered patentable and that this creation can be attributed to DABUS. However, the patent application must be formally filed by a natural person, so the procedure initiated by Thaler is not valid.

The only positive result was achieved in South Africa: the country’s patent office actually considered the request valid and therefore granted the patent for a new type of food container for AI designed by Thaler.

Thaler’s efforts, which have not yet been completed, have highlighted all the problems in the current regulation in the field of patents. The food container can be considered as a marginal detail, the focus of all these processes active in different parts of the world is the question posed at the beginning of this article: can an AI be considered an inventor in all respects?


Historically, law has found it difficult to keep up with technological innovation, and this is a brake on investment and thus research. Recent advances in machine learning have made regulations around the world seem outdated.

We need new definitions and new legal tools that are able to anticipate the future. Although we are still far from strong AI, that is, from AI that can mimic human thinking, we can already see how autonomous machines are beginning to question complex concepts of rights, duties, and even the very definition of legal personality, existence, and sensing.

We can say with confidence that Thaler’s struggle is a first step towards defining a legal framework that takes into account the current state of technology, but which, above all, prepares for what the future of computing will be.

Trends 2022 for UC&C: how to make business communication smart

Leave a Comment