artificial intelligence

This article is written by Harsha Aswani, pursuing a Diploma in IP, Media and Entertainment Laws from LawSikho. The article has been edited by Tanmaya Sharma (Associate, LawSikho).

This article has been published by Shoronya Banerjee.

“The new generations bring with them their new problems which call for new rules, to be patterned, indeed, after the rules of the past, and yet adapted to the needs and justice of another day and hour.”


Until quite recently, creation and innovation were exclusively associated with the human mind, and so our legal framework and various theoretical justifications were built around this very aspect, focusing on incentivizing or encouraging the highly intellectual and innovative species of the world to come up with new and better solutions to our problems. 

However, the exponential development in the field of artificial intelligence has not only revolutionised the globe but has caused a shift in the paradigm and made it seem imperative to redefine the interpretations of our existing framework of IPR laws.

This shift in the paradigm is largely attributed to that area of artificial intelligence technology which per se can generate newer inventions in varied fields with little or insignificant human intervention or exclusively by itself. It has changed the dynamics of the interface between AI and traditional IP Laws.

While the Patent Offices and the Courts of many countries such as the US, the EU, China, Australia, Africa, Japan have tried to interpret the existing patent laws to respond to the patent applications centered around AI being an inventor and has made some efforts to accommodate this massive breakthrough technology, Indian Patent Office or Courts have yet not taken any such initiative.

This article shall attempt to analyse the positions taken by EPO and USPTO concerning those patent applications embodying AI as an inventor, and will try and substitute the considerations to see how Indian Patent Act would respond to AI being an ‘Inventor’, and whether the standards of ‘Non-Obviousness’ and ‘Person Skilled In The Art’ (PSITA) need any amendment or new interpretation. 

In late 2018, Dr. Stephen Thaler filed two patent applications in the patent offices of three major jurisdictions, namely, the European Union, the UK, and the US, in his name as an “applicant” while designating DABUS, a “Creativity Machine”, as an “inventor”. The claims in one of the patent applications were largely related to a fractal design of a beverage container while the other application claimed a patent for a “neural flame” device used in search-and-rescue missions. The major contention put forth by Thaler in his application, which draws a line between AI-assisted and AI-generated inventions, was that the machine called DABUS has, without any human intervention, not only conceived the present inventions but it has been able to appreciate, on its own, that it had an invention, which was novel, non-obvious and useful.

While Thaler did not provide how DABUS appreciated that he had a patentable invention, nor was this challenged by anyone, but Thaler did clarify that the machine was fed only the data related to the general knowledge of the field for training, distinct from the field of the present inventions and yet has independently been able to devise and generate the potential solution in response to the unspecified problem.

The team “Artificial Inventor Project”, which is behind DABUS and similar technology, does not say that a patent should be owned by an AI, but only that it should be identified as an inventor, for two-fold reasons:

  1. That hundreds and thousands of employees have contributed code to such a system, as such patent protection over such innovations generated by AI will enable all such humans to reap benefits and recoup the investment made in Research & Development over the years.
  2. That it would promote creativity as it would be more explicitly known for the values such AI systems add.

In response to the respective patent applications listing DABUS as the only inventor, all the three patent offices− EPO, UKIPO, and USPTO, have refused the application being invalid. Even during the appeal preferred by Dr. Stephen Thaler in the UK High Court, Justice Marcus Smith ruled that the application is not a valid application, despite the invention meets the three qualifiers of patentability, for a Machine Learning Model cannot be listed as an “Inventor”. The rationale behind the decision was straightforward, i.e. the inventor must be identified as a natural person such that the designation of inventorship as prescribed by UK Patent law must contain “a family name, given names and full address of the inventor.” Merely giving a name to the machine, DABUS, does not identify ‘a person’ required by law. Further, Marcus J. also pointed out that since, currently, there is no such legislation or jurisprudence that accords rights to a Machine Learning Model so they have no legal personality comparable to natural or legal persons, as such DABUS cannot transfer/assign those rights that he is not himself entitled to hold in the name of Dr. Stephen Thaler, who is listed as an ‘applicant’ in the Patent Form 7.

Similarly, the USPTO while refusing to accord patent protection to DABUS ruled that the US Patent Law defines an ‘inventor’ as an ‘individual’ or ‘individuals’ (in case of Joint inventorship) and consistently uses human-like words such as “whoever” and pronouns like “himself” and “herself” to refer to ‘inventors’, which reinforce the idea that only ‘natural persons’ may be named as an ‘inventor’ in a patent application, and it shall be too broad an interpretation to include machines thereunder. In addition to it, the USPTO relied on a decision of the Federal Circuit to explain that US Patent law requires only a natural person to be an inventor. It states, 

“Conception is the touchstone of inventorship, the completion of the mental part of the invention. It is the formation in the mind of the inventor, of a definite and permanent idea of the complete and operative invention, as it is hereafter to be applied in practice. Conception is complete only when the idea is so clearly defined in the inventor’s mind that only ordinary skill would be necessary to reduce the invention to practice, without extensive research or experimentation. Conception is a mental act. …”

The takeaway from this decision is that one, Conception is an inherently human activity, and two, such conceived invention must be reduced to practice by such a person who conceived it without any undue experimentation or research. The first part “Conception” of an idea is what creates a major hurdle for AI does not seem to have a mind of its own, and it is believed that the output is majorly the product of the input, in the form of data, fed into the AI system. But what if the output that is generated by AI is nowhere related to the input fed into it? Does it suggest that AI has a mind of its own, or if not a mind of its own, but can the self-learning ability of an AI be held responsible for such conception? Is this self-learning ability due to the algorithm designed by the developer? If that is so, won’t he be the rightful inventor or can he be given the ownership of such a patentable invention along with holding AI as an inventor? 

The decisions above, instead of settling down the uncertainty surrounding the Patenting of an AI, have opened a Pandora’s Box filled with many significant questions that can no longer be ignored. The question substantially relate to –

  • Should the law permit or require that the AI model be named as an ‘Inventor’ or should it be required that a human being be named as the ‘Inventor’; or
  • For the purposes of naming a human being an ‘inventor’ for an invention which was perhaps majorly conceived by the AI, what level of involvement is needed by a human, or what level of human interaction is needed to list that human as an ‘inventor’, in addition to or in place of an, AI such that the application is deemed to be valid; or
  • If AI is named an ‘Inventor’, then will it bear the burden to prove that it was the first to conceive as well as appreciate the invention; or
  • How would an AI even go about appreciating that it has an invention; or
  • What happens if the human steps in to recognize the conceiving of the invention by AI, and says, “Hey, this is an invention”; 
  • What would it take for an AI to be an ‘inventor’; 
  • If the machine enjoys the legal status of the inventor, will such the machine be able to be held liable for infringement; 
  • Even if a law permits an AI to be named as an ‘Inventor’ and grants the patent to inventions autonomously generated by AI, how would then the standard of the hypothetical person skilled in the art be set for determining the prior art and the non-obviousness of the future inventions, whether created by humans or by the AI? 

These are a few of the questions that naturally fall into the lap of a human being while dealing with the patentability of an AI-generated invention.

 What happens after DABUS?

Seeing the plethora of uncertain questions hovering around the patenting of inventions autonomously generated by DABUS and similar technology, the EPO and the USPTO have laid down following comprehensive guidelines as to what extent AI-related inventions can be protected by patents.

European Patent Office (EPO)

Every year, new guidelines are furnished by the EPO that demonstrate the current practice of patenting inventions undertaken by it. This time, in November 2019, EPO made some substantive changes to accommodate the AI-based inventions, including the Machine Learning (ML) Models, by granting them the status of technical solutions, which is a prerequisite for recognizing an invention as such.

AI and Patentability

With specific reference to the Patenting of the Fourth Industrial Revolution, initiated due to the exponential advancement in the field of Artificial Intelligence, the earlier approach adopted by EPO was that it considered AI-based inventions to be of non-technical nature. However, the latest 2019 guidelines issued by the EPO, suggest that AI-generated inventions may be assumed as having a technical character, provided they pursue a technical purpose.

For instance, classifying text documents solely in respect of their textual context is not regarded to be per se a technical purpose but a linguistic one.

Complying with Article 27 of TRIPS Agreement, Article 52 of EPC, para 1 provides that the patent shall be accorded to all inventions, in all fields of technology, without any discrimination, provided they are novel, have an inventive step, and are susceptible of industrial application. However, para 2 explicitly keeps certain inventions out of the patentable subject matter, including programs for computers. Thus, core AI, for example, algorithms as such, are not patentable whereas AI-generated inventions might be eligible for patent protection if they have features, realized wholly or in part, that show sufficient technical character.

Hence, AI-related inventions requesting for the grant of the patent must conform to the following requirements:

  1. The invention must have a technical character. It implies that such an invention must-have features that contribute to the solution of a technical problem by providing a technical effect.
  2. The technical features of an invention must involve an inventive step. It implies that the technological advancement made over such an invention, as compared to the prior art, must be not obvious to the hypothetical person skilled in the art.
  3. The claims and description must be clear and devoid of any marketing terms, slang or jargon.
  4. The application must be disclosed sufficiently to enable the hypothetical person skilled in the art to implement the patented invention without undue experimentation, after the patent term expires, merely by reading the claims and the description.

Inventorship of AI Inventions

The extraordinary advancement in the field of artificial intelligence and machine learning has initiated a debate all around the globe, forcing the G20 nations to embark on a journey to curate, apply and find the leveragability of artificial intelligence. The debate has stirred various views, the significance of which is that AI must be acknowledged as an “inventor” since it has the capacity to invent just as a ‘human’.

The EPO, for clarification, has divided AI inventions into three broad categories from the perspective of ‘inventorship’:

  1. Human-made inventions using AI for the verification of the outcome;
  2. Inventions in which a human identifies a problem and uses AI to find a solution;
  3. AI-made inventions, in which AI identifies a problem and proposes a solution without human intervention.

The first two categories fall within the domain of AI-assisted inventions where AI is used as an inventing tool for human inventors, augmenting their capabilities. This domain of AI-inventions does not create any grave issues as such, as against the third category which is that of AI-generated inventions, whereby AI, autonomously and independently invents without any human intervention in the form of any direction, instruction, and oversight. This very area of inventions is what we have always read and heard about in science fiction, and has both its pros and cons that need to be approached in a healthy and balanced fashion.

So far the common understanding that prevails in the EPO, and as confirmed by the EPC, is that the inventor can only be a human being and not the machine and that such a person must create the invention by their creative activity. The view is supported by the rationale that the designation of an inventor attracts a variety of legal implications ensuring that the named inventor is a rightful and a legitimate inventor, who can, in practice, reasonably exercise the rights and privileges attached to such a status. Since AI systems or machines do not have a personality recognized by law, they cannot enjoy any such benefits accruing therefrom.

United States Patent and Trademark Office (USPTO)

Clause 8 of Article 1 (8) of the US Constitution empowers Congress to promote the progress of science and useful arts by way of securing exclusive rights of the authors and inventors concerning their respective writings and discoveries, for a limited period. Maintaining the foregoing object laid down by the framers of the American Constitution, especially concerning the emerging technology such as Artificial Intelligence, the USPTO sought feedback from various stakeholders and experts and had issued in the latter half of 2020, certain guidelines about this breakthrough technology, which is majorly a compilation of the numerous comments that it received on some significant issues related to patenting of AI inventions. 

AI and Patentability

The report defined an AI invention to be a “computer functionality that mimics cognitive functions associated with the human mind (e.g., the ability to learn).” The USPTO, for now, considers AI inventions as part of Computer-Implemented Inventions, as such, they do not possess any unique patent eligibility considerations which may be otherwise specific to AI inventions. Therefore, they are treated like any usual invention that comes before the US Patent Office. Apart from this, the USPTO also divided the AI inventions into the following broad categories:

  1. Inventions that embody an advance in the field of AI (e.g., a new neural network structure of an improved machine learning (ML) model or algorithm);
  2. Inventions that apply AI (to a field other than AI);
  3. Inventions that may be produced by AI itself.

Largely the commentators were of the view that such AI-generated inventions are more likely to be classified as abstract ideas, or mathematical concepts, or mental processes, thereby attracting the subject-matter exceptions created judicially. However, one commentator noted that complex algorithms substantiating an AI invention can generate an output having some technological improvement as compared to the prior art, and therefore, may be patent-eligible. Additionally, the report suggests that even the claims directed to abstract ideas may be patent-eligible if the additional claim elements, whether individually or as a combination, show significant transformation as compared to the earlier abstract idea.

 AI and inventorship

The development of an AI system is usually attributed to a large number of innovators, as such the law requires careful determination of the identity of such a person who has, in the true sense, contributed towards the conception of the AI invention to be legally called an “inventor”.

35 U.S.C. § 100 defines an “inventor” as the “individual” or “individuals” (in case of joint inventorship) who invent or discover the subject matter of the invention. Further, 35 U.S.C. § 116 explicitly lays down that an invention may be made by two or more persons jointly despite:

  1. They did not physically work together or at the same time;
  2. Each did not make the same type or amount of contribution; or
  3. Each did not contribute the subject matter of every claim of the patent.

An additional clarification is afforded by 35 U.S.C. §115 which provides that an Inventor must only be a natural person, for the words “whoever”, “himself” or “herself” can be attributed to a human being only, who shall make an oath or declaration that such a person believes himself/herself to be an original inventor of the claimed invention.

Moreover, the Federal Circuit has also clarified and explained the importance of “conception” as the “touchstone of inventorship” and has affirmed time and again that inventorship can be attributed to such person in whose mind “the formation of a definite and permanent idea of the complete and operative invention” takes place.

Since the US Patent Law doesn’t recognize the machine or a company as an “inventor”, so one could infer that the existence of human contribution is mandated both by the legislator and the judiciary and that the amount of contribution required by a human to validate the application for grant of a patent shall be evaluated on a case-to-case basis. A similar line of argument was put forth by even one of the commentators in the report compiled by the USPTO.

The majority of the commentators negated the idea of revising the US Patent Laws and regulations, for now, to include machines or any entity or entities other than a natural person as an “inventor” because first, “conception is inherently a human activity” such that an entity other than the human being could not, in a real sense, contribute to the conception of an invention; and second, “the current state of AI technology is not sufficiently advanced at this time and in the foreseeable future to completely exclude the role of a human inventor in the development of AI inventions.”

Patenting of AI inventions in India

India is emerging as a new destination for the filing of patent applications involving AI. In the year 2015, India was ranked eighth for first filings and has enjoyed a high rate of annual growth, with an average of around 33 percent in the 3 years up to 2015. However, despite being one of the predominant choices for filing patent applications in the field of AI-driven technology, the Indian Patent Office has maintained its silence on the issue of patentability of such inventions. 

AI and Patentability

To become TRIPS-compliant, Section 3(k) was introduced in the Patents (Second Amendment) Bill, 1999, which explicitly categorized mathematical methods, computer programs, and algorithms as non-patentable subject matter. The bill was then referred to the Joint Parliamentary Committee, which recommended adding the term ‘per se’ as a suffix to a computer program. 

The term ‘per se’ has not been defined anywhere in the Indian Patent Act, and therefore, it has been a subject that is debated too often. The views expressed by the Joint Parliamentary Committee during the introduction of the 2002 Amendment, however, attempts to clear the intent of suffixing the word ‘per se’ to computer programs as:

“In the new proposed clause (k), the words ‘per se’ have been inserted. This change has been proposed because sometimes the computer program may include certain other things, ancillary thereto or developed therein. The intention here is not to reject them for the grant of the patent if they are inventions. However, the computer programs as such are not intended to be granted patents. This amendment has been proposed to clarify the purpose.”

This section does not, however, impose a blanket ban on patenting of Computer-Related Inventions (CRI). Only inventions that disclose computer programs per se are not regarded as inventions under the Indian Patent Law. 

There have been debates about treating AI-based inventions similar to Computer-Related Inventions (CRI) which include software programs because, at its core, such AI systems use a complex set of algorithms that enable them to achieve the intended result. This approach is similar to that of the EPO, which requires an AI-based invention to have some specific technical purpose, i.e., it must demonstrate some specific use, and specify the technical effect of the use of the algorithm, to be patent-eligible; otherwise, such inventions shall fall within the domain of non-patentable subject-matters as per the EPO. 

Similarly, the guidelines for examination of CRI issued by the Indian Patent Office also demand that for such inventions to be patentable, they must possess some technical character, which shall be determined by the substance or the nature of the claims. The claims must sufficiently demonstrate that such inventions have some ‘technical effect’ and have made certain ‘technical contributions’ to the prior art, to be patentable. In other words, it must solve a technical problem, and the method claim should clearly define the steps involved in carrying out the invention.

“The focus should be on the underlying substance of the invention, not the particular form in which it is claimed.”

Under Section 3(k) and the Guidelines concerning Computer Related Inventions (CRI), the single judge bench of the Delhi High Court, in 2019, while adopting a liberal approach, remarked that the IPO’s guidelines are too restrictive in light of the technological advancement as well as the global practice adopted worldwide. The court observed:

“The bar on patenting is in respect of ‘computer programs per se…’ and not all inventions based on computer programs. In today’s digital world, when most inventions are based on computer programs, it would be retrograde to argue that all such inventions would not be patentable. Innovation in the field of artificial intelligence, blockchain technologies, and other digital products would be based on computer programs, however, the same would not become non-patentable inventions simply for that reason. It is rare to see a product that is not based on a computer program. Whether they are cars and other automobiles, microwave ovens, washing machines, refrigerators, they all have some sort of computer program in-built in them. Thus, the effect that such programs produce, including in digital and electronic products is crucial in determining the test of patentability.”

This judgment attempts to clear the air around the patentability of computer-related inventions which are made up of complex algorithms.

However, the optimization of the heuristic algorithms has lessened scope for human intervention in the creative or the development process of an invention involving AI, and the identification of an actual inventor has become a major hurdle. The question boils down to the fact− whether an AI system that has, because of its self-learning abilities, itself developed an invention that meets the threshold of patentability, could be called an actual ‘inventor’, or is the title of inventorship only related to a human being? What happens when all the persons associated with an AI system cannot reasonably say that they have contributed towards the conception of such an invention and have been able to reduce it to a working invention?

AI and inventorship

While the EPO and USPTO have made it amply clear that the legal definition of an ‘inventor’, for now, is primarily restricted to including a ‘natural person’, India has not made any such clarifications officially. Moreover, the Indian Patent Act does not explicitly define the term ‘inventor’, unlike EPC and Title 35 U.S.C, so the question of restricting the title of inventorship to only natural persons does not seem to be the only approach. 

Article 27(1) of the TRIPS Agreement which provides only the criteria for qualifying an invention is patentable, does not explicitly state that if the invention, whether products or processes, must be made by humans or may also be made by an AI system. In other words, TRIPS, in principle, does not explicitly restrict the right of inventing to humans only in any explicit terms. 

In the Indian context, Section 6 (1) of the Patent Act requires that the application for the grant of a patent may be filed by any ‘person’ who is the ‘true and first inventor of the invention or the ‘assignee’ of the true and first inventor. Further, section 2(1) (p) also defines a ‘patentee’ as a ‘person’ entered on the register as the grantee or the proprietor of the patent. One may assume that an ‘inventor’ or a ‘person’ only includes a ‘natural person’. 

However, the term ‘person’ as defined in section 2 (1) (s) includes ‘government’, which is a non-human entity. Since the definition is inclusive, it may be broadly interpreted to include an ‘artificial person’ or a ‘non-human entity’. 

Clause (42) of section 3 of the General Clauses Act, defines a “person” to include a company or association or body of individuals, whether incorporated or not. The term ‘shall include’ makes it clearer that the term ‘person’ is not merely restricted to include a natural person, and may include an artificial person too.

Even the term “true and first inventor” as defined in section 2(1)(y) is merely an exclusionary definition and makes no mention of a natural person as such.

The only difference concerning government, as a non-human entity or the ‘company, association or body of individuals as a non-human entity, on one hand, and AI system, on the other hand, boils down to the fact that law has attributed legal personality to the former category of non-human entity, which is why they are regarded as ‘Legal Persons’, while this isn’t the case concerning AI. The rationale behind attributing legal personhood to the above category of non-human entities is that the decision-making power behind such entities vests in the humans, ultimately, which is understandably not the case with AI. 

So, before attributing ‘inventorship’ to AI, the question of attributing legal personality must also be answered objectively, after weighing the pros and cons of the same, in light of all the social, moral, and societal implications. Recently, ‘Sophia’, a social humanoid robot developed by Hanson Robotics, a Hong-Kong based company, has been granted full citizenship by Saudi Arabia. The question now pertains, which is for the legislature to decide, if Sophia after being given citizenship may be considered a natural person or not, and the implications that may follow from it.

Also, the further argument against it could be from the point of ‘conception’ of the invention, which is the ‘touchstone of inventorship’. AI system, being the complex set of algorithms, and involving multiple people behind its development, the question of inventorship shall again fall into the trap of ambiguity, especially when it is not known who is responsible for the conceiving of the invention, at first place− the AI itself (using its self-learning ability) or the human who developed the algorithms for such AI that it can self-learn, or the person who feeds the data as an input to the AI, or the one who reviews the result and recognizes that an invention has been made.

This may be determined once the legislature lays down sufficient guidelines to set the required threshold limit of the amount of contribution required by the human inventor to be named as an ‘inventor’ in the patent application along with an AI or place of an AI. The idea behind attributing inventorship is to acknowledge the contribution made by the persons involved in developing such an invention. If such an invention is made by an AI, the law can acknowledge the contribution of AI by way of the “right of attribution that exists vis-à-vis the inventor.”

“The right of attribution for an AI which is envisioned forms a part of a pseudo-patent application wherein the AI involved is named. Such recognition should be conditioned on the extent of real intellectual efforts towards developing the invention itself which was not initially perceived by the inventor himself. Such a right might seem rather useless for the AI, however, it can be helpful when− one, it provides traceability to the extent of liability that the human inventor may be awarded; and two, buried in the right of such attribution to AI is the repute of the human inventor involved which would not die out even after assignment thus, providing a flexible degree of control over one’s creation. The pseudo-patent application is a superficial patent application, supplementary to the main patent application. The details regarding the extent of intellectual efforts by the human inventor and the AI involved should be recorded in the pseudo-application. Since the inventor cannot manually distinguish for an AI which logic is wasteful and should be avoided, the AI sans discernment reflects the system biases thus, inflicting liability on the inventor or even the corporate assignee.”

So far, the Indian Patent Act does not conclude that only human inventors should be named as ‘inventors’. In case, the legislature desires to prohibit an AI to be named as an inventor, it would require amendments in the existing patent law to explicitly state the same.

AI and ownership

The question of ownership, in the case of AI-generated inventions, is far more pressing than that of inventorship, given the legal implications and the rights that follow. Ownership in AI-generated inventions will encourage and push the persons behind the development of the AI system towards further innovation and shall enable the businesses to invest more in the technology leading to better commercialization of the invention.

Generally, an inventor is deemed to be the first owner of the invention. However, it is amply clear that for now AI cannot be deemed to be an owner of the invention. Thus, to determine the rightful ownership of AI-generated inventions, it is imperative to determine, first, the actual inventor of the invention, as without acknowledging who the ‘true and first inventor is, it is not possible to identify who owns the invention. This sequence is important and cannot go the other way round especially concerning AI inventions as it is argued that an AI system cannot be named as an ‘inventor’ because it can neither exercise the ownership rights in the patent nor can it assign something which it does not have in the first place, for it lacks legal personality.

One approach could be simply to attribute ownership to the inventor of an AI system (hardware) that generates ideas and makes the patentable invention. Another approach could be to grant ownership to the developer of the algorithms (software) incorporated into such an AI system where it is argued that it is the input that is responsible for what the AI generates as an output. This shall not only act as an incentive for such people to innovate further, but it may also act as a self-regulatory check on the AI and will hold such inventor or the developer morally as well as legally responsible to keep the AI within the bounds, at least, up to the initial phase where such human can exercise sufficient control over the output an AI generates. Although some argue that it isn’t correct to make the innocent human liable for things that might go wrong with AI and that it violates human rights as well, then the line has to be drawn that shall, at least to some extent, if not absolutely, hold the human behind such AI responsible, even if that can be limited to only the initial phase of control. This may be regulated by certain contractual agreements laying down specific terms and conditions, and setting some bounds and limitations which may be subject to judicial review.

Further, this may also ease the right of a human inventor to assign or license the right to practice such AI-generated inventions to any person he deems fit in the same manner as it is usually done in the case of other patentable inventions. 

AI, non-obviousness, and the person skilled in the art 

The non-obviousness of an invention is the second qualifier of patentability. It is a subset of inventive steps and is determined from the perception of a hypothetical person, known as “Person Skilled in the Art” (abbrev. as PSITA), in the context of India.  

Section 2(1) (ja) defines inventive step, which has two parts:

Technical advancement or economic significance or both

The features claimed in the invention must show sufficient technical advancement when compared to the prior art. In other words, such an invention must add some value, which is of technical nature, to the existing prior art; or such an invention must be economically viable as compared to the existing prior art, making the invention more affordable by utilizing machinery to help the public to have more access to the invention.


Such technical advancement or economic significance must be non-obvious to the hypothetical PSITA. The PSITA places the prior art alongside the claimed invention to identify the difference between the two. If he is unable to see through any difference as a whole, after mosaicing all the prior art documents, before the effective filing date of the application, then the invention shall be deemed to satisfy the test of non-obviousness, and consequently, be deemed to have an inventive step. Under AI-generated inventions, the question largely revolves around two major issues:

1. Should the standard of the hypothetical PSITA be maintained where the invention is autonomously generated by an AI? What shall be the probable implications if AI is elevated to the position of PSITA?

In the context of AI inventions, the determination of who such a hypothetical person is shall be a major question− the AI or the human contributor. Although generally, the hypothetical person is a construct of legal fiction, which possesses average or ordinary knowledge of the prior art of the relevant field such that when he looks at the claimed invention, he must not be in a position to carry out such an invention without undue experimentation. He is not deemed to possess the knowledge of the inventor, which makes the claimed invention non-obvious in the eyes of PSITA. 

However, with the rapid advancement in technology, one fears the time when such AI shall become super-intelligent and all the future inventions shall start to become obvious for the ‘inventor’ AI, which might demand recognizing AI as a hypothetical PSITA. However, elevating AI to the position of PSITA has its consequences that make this approach less feasible in the light of future innovations, particularly those coming from the end of humans. Placing AI in the position of PSITA shall set the patentability standard too high and will act as an impediment to the deserving innovations generated by humans. Theoretically, this might seem like a bar to junk patents, but practically, the risks seem to be too high to even consider taking such a step. It shall force redefining of all the existing qualifiers of patentability, including novelty, enablement, and the disclosure requirement; particularly enablement and disclosure requirement which is viewed from the standpoint of the person skilled in the art.

2. What art does the standard refer to?

Since AI-generated inventions belong to not one but a combination of fields, the question as to what the prior art standard should be limited to demands attention. Given the versatile and continuously evolving character of AI, the definition of ‘AI-invention’ must be first articulated in a well-defined manner to determine the standard the prior art must refer to. For now, owing to so much uncertainty around the AI, it seems feasible to restrict the standard of prior art to the field to which the claims invoked essentially relate, rather than establishing any peculiar standards of legality for particular technical art.


Artificial Intelligence is here to stay. It is not even a question anymore if inventions involving AI technology should be patentable or not. Not affording patent protection to AI-driven inventions would force the applicants to conceal the involvement of an AI, which has been a predominant practice so far in this area of technology. Apart from this, exclusion of AI-based inventions from the patentable subject matter will lead to an increase in the use of trade secrets, which is any day better, from the standpoint of the patent applicant, as compared to allowing the invention to fall into the public domain. The ramifications of such practice will decrease the flow of information and technology advancement, as such, it is in the interest of both, the applicant as well as the legislature and the Indian Patent Office to start reacting to the developments within the field of AI technology by way of either enacting appropriate laws and policies, or by interpreting the existing laws to accommodate such inventions, and if required, introducing the amendments for the same.

Further, the question of inventorship and ownership are important not only from the perspective of acknowledging and encouraging the ‘person’, whether natural or artificial, behind the patentable inventions, but also for commercialization of the invention, and analyzing the more pressing issues linked to it, including statutorily mandated standard of hypothetical PSITA, the sufficiency of disclosure, infringement, liability, and dispute resolution. 

As discussed above, the question of identifying PSITA is important to restrict its use exclusively to the statutorily appointed role during the non-obviousness determination, and not for other purposes, like enablement. The question of Enablement poses another level of difficulty, for instance, will AI be able to, at all, disclose the process of bringing about an invention in the patent specification? Will the Human owner, who is not involved in the process of making the AI, be able to back-track and reverse engineer the same process to produce the definite output, if at all disclosed by AI, by looking at such specification? Wouldn’t the “inherent randomness in AI-algorithms” make it difficult for such a PSITA to essentially understand the functioning of the invention, and reverse engineer the same without undue experimentation? If at all it is possible, the other questions will then become more difficult to answer, like if AI is granted the status of inventorship, would it be able to bear the burden of proving that it was the first to come up with such invention, in case of any conflict between the title of inventorship between two separate inventions made by separate inventors, one of them being the AI? Will the litigators be able to question the AI, during depositions, to explain the logic behind the concerned invention?

Recently, in 2021, the Australian court and South African court have ruled in favor of recognizing AI as an inventor, but the above issues still remain unanswered. The timely recognition and response towards such issues is the only way forward to keep pace with this multifaceted and evolving AI technology.


  2. (Creativity Machine is a type of Connectionist AI which generates intelligent behavior by representing rules in interconnected networks of simple and uniform units such as artificial neurons, and includes Machine Learning, Artificial Neural Networks and Deep Learning.) 
  3. WIPO Conversation on IP and AI, second session, Draft issue paper on Intellectual Property and Artificial Intelligence, 5, (last visited Jan.3, 2021).
  4. Jon Porter, US patent office rules that artificial intelligence cannot be a legal inventor, THE VERGE (Jan. 1, 2021, 1:00 PM),, see id. (RYAN ABBOTT (Artificial Inventor Project) suggesting that recognizing the value of machines in the creative process will inevitably make the machine become more valuable).
  5. Univ. of Utah v. Max-Planck-Gesellschaft zur Forderung der Wissenschaften e. V., 734 F.3d 1315 (Fed. Cir. 2013).
  6. EUROPEAN PATENT OFFICE, Patents and the Fourth Industrial Revolution (Guidelines, Part G, Patentability, Chapter II- Inventions, 3.3.1. Artificial Intelligence and Machine Learning), (last visited Jan.1, 2021).
  7. Agreement on Trade Related Aspects of Intellectual Property Rights, Apr. 15, 1994, Marrakesh Agreement Establishing the World Trade Organisation, Annex 1C, art. 27, (hereinafter TRIPS).
  8. USPTO, Public Views on Artificial Intelligence and Intellectual Property Policy, I, 1 (2020),

Students of Lawsikho courses regularly produce writing assignments and work on practical exercises as a part of their coursework and develop themselves in real-life practical skills.

LawSikho has created a telegram group for exchanging legal knowledge, referrals, and various opportunities. You can click on this link and join:

Follow us on Instagram and subscribe to our YouTube channel for more amazing legal content.


Please enter your comment!
Please enter your name here