Readers may recollect that the authors of this article had previously published "The Practitioner's Dilemma of Patenting AI in India". This is a continuation, however, herein we discuss the unending conundrum faced while patenting AI related inventions in India, under the Patents Act, 1970 ("Act").
Challenges continue to lie in respect of inventorship, ownership, eligibility, inventive step and adequacy of disclosure. However, there is still no guidance from the office of the Controller General of Patents and Designs ("IPO"), which furthers complications. This is set against the growing number of patent filings in this technical field. Suitable amendments to patent policy are now required more than ever before. Authorities must act with urgency to create a clear and predictable path, failing which, they risk disincentivizing development of path breaking innovations.
A round table discussion, or draft guidelines would be welcome.
When faced with such applications, the IPO continues to examine them as CRIs in accordance with the 2017 CRI Guidelines ("Revised Guidelines") which instruct examiners to focus on the underlying substance of the invention over form of claims. In fact, an analysis of approximately 80 cases examined in AY 2019-2020, confirmed a consistent pattern among examiners to treat AI related inventions as computer programme per se or algorithms and therefore object under Section 3(k) of the Act, which excludes mathematical or business methods or a computer programme per se or algorithms, from eligibility.
Recently in Ferid Allani vs Union of India &
Ors.1, the Delhi High Court while reiterating the
principles discussed in the 2013 Draft CRI Guidelines ("Draft
Guidelines"), unsurprisingly held that CRIs are not barred
from patentability, but that such an investigation requires
assessment of the "technical effect" or "technical
contribution" of the underlying invention. This decision was
passed by the Court while exercising its writ jurisdiction, which
was invoked to challenge the decision of the Intellectual Property
Appellate Board, which had confirmed the IPO's decision to
refuse the Petitioner's patent application "a method
and device for accessing information sources and services on the
web". The Petitioner argued that "the
specification disclosed a technical effect and technical
advancement, and that the claims did not relate to a mere
software but a particular method of
implementation." Allowing the petition, the
Court held that "the bar on patenting is in respect of
`computer programs per se....' and not all inventions based on
computer programs" and noted that "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. ..........
Thus, the effect that such programs produce
including in digital and electronic products is crucial in
determining the test of patentability." The
Court further held that "patent applications in these fields
would have to be examined to see if they result in a
`technical contribution‟" and that
"If the invention demonstrates a "technical
effect‟ or a "technical contribution‟ it is
patentable even though it may be based on a computer
program."
This ruling among other things has burst wide open the door to patentability of AI related inventions. While this decision tips the scales in favour of allowance, it has, understandably, not addressed finer aspects, which, as we discuss in this article, need addressing to fully realize the possibility of obtaining patents for AI related inventions.
INVENTORSHIP AND OWNERSHIP
First, is the widely debated issue of a machine being permitted or rather required to claim inventorship. The Act does not explicitly define the expression "inventor", and instead, permits inter alia the person(s) claiming to be the true and first inventor, and the assignee of such "person(s)", to file a patent application. The definition of the word "person" and the expression "true and first inventor" do not advance the position. Therefore, in our view, the Act cannot be relied on to answer this issue. The cases of V.B Mohammed Ibrahim vs. Alfred Schafranek & Ors.,2 and Shining Industries vs. Sri Krishna Industries3 provide some clarification albeit only while determining whether a legal person or a financial contributor can claim inventorship. To both issues, the respective courts ruled that an inventor must provide technical contribution and be a natural person. In another case National Institute of Virology vs Mrs. Vandana S. Bhide4, the IPO ruled that a pair of hired hands cannot claim inventorship, and to be named as an inventor, such person must have made some intellectual contribution in achieving the final results of research work leading to a patent. Quite understandably, none of these cases contemplated an inventorship claim by a third category of person i.e. an artificial one. However, if one was to apply the ratio of these decisions, it may not be wholly incorrect to say that a machine must be recognized as an inventor, if it provides a technical contribution which is material to patentability. It is also well understood and acknowledged that it would be erroneous to omit the machine from being named as an inventor. Discrediting a machine is not really a matter of concern, but unduly crediting a human may affect the integrity of the entire patent system. Hence, the question remains. Some of these issues are currently being debated aggressively pursuant to the DABUS decisions from the USPTO and EPO.
The issue of inventorship also generally affects two other essential rights that lie in the underlying invention, namely, the right to apply and own. Under the Act, any person (i.e. natural or legal), other than the named inventor, who applies for a patent, must prove that he is entitled to do so. This, in practice is referred to as the "proof of right". To fulfil this requirement, applicants can either provide a written deed of assignment or an endorsement by the inventors on the prescribed Form 1 of the Act. Further, ownership of the right, title and interest, in and to the underlying invention, patent application and patent is ordinarily transferred by means of an assignment. Under Indian law, for an assignment deed to be a legal document, it must satisfy the essentials of a valid contract, be reduced to writing and duly executed. These include offer-acceptance, lawful consideration, free consent, capacity to contract, and lawful objective. In some jurisdictions in employer-employee relations there is an implied assignment, without the need for a specific document, however, this is not the case under Indian law, which requires that such assignment be expressly provided for in the employment agreement. Under the current framework, there is a need to clarify whether machines are competent to contract, what could be lawful consideration, and how one could establish mutual assent. If this is not clarified, it is safe say there is a great burden on the applicant-owner to establish its right to apply and title.
PATENTABILITY
The IPO's treatment of AI related inventions as CRIs and objecting under Section 3(k) is confirmed by a cursory review of recently examined patent applications directed to AI related inventions. This is because there are no specific provisions governing their eligibility and due to limited discussion on the need for applying a different criterion.
Then, the instructions in the Revised Guidelines are exemplified by saying that the mere presence of a mathematical equation in a claim would not render it a mathematical method. For instance, the mere presence of mathematical formula in claims directed to improving communication systems, would not cause them to be considered as mathematical methods. It is evident from patent applications published in India that AI related inventions broadly cover mathematical models as well as advanced mathematics, which is non-patentable subject matter. However, as confirmed in Ferid Allani (supra), they may qualify, if the claims include a technical means or are implemented by using a computer and/or are running on a machine, to solve a specific technical problem and producing a "technical effect" as per the Draft Guidelines of 2013. Interestingly, this aspect is not discussed in the Revised Guidelines, and therefore clarity in respect of the expression "technical" or "technical effect" is still lacking, which provides ample room for interpretation.
Next, computational models and algorithms are generally considered to be of a mathematical nature and therefore not patentable. However, it is overlooked that such models may contribute to the technical character of an invention, by producing a technical effect that serves a technical purpose, by: i) its application to a field of the invention, and/or ii) being adapted to a specific technical implementation. Thus, if a computational or mathematical model or an AI algorithm fulfils these and other criterion, and depending on the manner in which it is claimed, should not be rejected as being ineligible.
INVENTIVE STEP
The discussion on inventive step viz. AI related inventions has to be seen at least from two broad perspectives. One, where AI contributes to the inventive aspect of the invention, and the other where the AI is the "inventor,".
As discussed, since the self-inventing ability of a machine, may not have been considered during enactment, there is no guidance on how inventive step might be analysed where an invention is wholly created by an AI or where the AI components or module contribute to its essential features.
Section 2(1)(ja) defines inventive step as "a feature of an invention that involves technical advance as compared to the existing knowledge or having economic significance or both and that makes the invention not obvious to a person skilled in the art."
The examination of inventive step has been largely developed on judicial expositions. In F. Hoffmann La Roche vs Cipla Ltd 5 . after referring to Bishwanath Prasad Radhey Shyam vs Hindustan Metal Industries Ltd 6 and other foreign rulings, the Court laid down a five-step test for such an inquiry and further stated "whether an invention involves novelty and inventive step or is obvious is a mixed question of law and fact, and must depend on fact and circumstances of each case, and that though no absolute or uniform formula can be laid down to ascertain obviousness..........". The Court also identified the skilled person "a person who practices in the field of endeavour, belongs to the same industry as the invention, possesses average knowledge and ability and is aware of what was common general knowledge at the relevant date".
It is indisputable that identification of the skilled person in this field is a major bone of contention globally. And no one disagrees that human intelligence and creativity can be supplanted by AI. Therefore, setting the foundation of an inventive step inquiry becomes a complex task.
Further, current law generally limits the scope of relevant art to analogous prior art and reliance is made on this to argue against combinations accused as rendering the claims obvious. However, with an "Inventive AI", the scope and content of prior art may be expanded to go beyond analogous art. Due to the wider use of AI technologies, the prior art base is also likely to increase exponentially, as it can be argued that AI is capable of knowing and processing virtually everything, at least, all that is easily available. At the same time, the skilled person might be using AI increasingly and, consequently, the test of inventiveness might need reconsideration
SUFFICIENCY OF DISCLOSURE
According to the Section 10(4) of the Act, a disclosure is considered sufficient inter alia when the complete specification contains an adequate written description which clearly and unambiguously describes the invention, meets the best method requirement, and allows a person of average skill and knowledge in India to perform the claimed invention without undue experimentation. This has been reiterated in several decisions including Press metal corporation limited vs Noshir Sorabji Pochkhanawalla 7 , where the Court ruled that "the invention is described in an obscure and ambiguous language and on this ground, the patent is liable to be refused."
However, as is often seen in AI related inventions, the input and output are known, but the logic, which may involve an AI-tool, in-between, is not. Here, even if the decision process is described, the execution of the same does not guarantee the same result. This is one of the precise reasons why AI related inventions can be difficult to disclose adequately.
Other problems may arise where an invention relying on the execution of a mathematical model is claimed, but the specification does not clearly teach how said model is brought to a working implementation. For example, if the model is implemented in the form of a neural network, it may be necessary to describe in detail the network topology and how the weights are set. If the inventive contribution is in the algorithm, the latter must be disclosed, but not, if it lies in the use of data. Furthermore, a mathematical model may contribute to the technical character of an invention by producing a technical effect that serves a technical purpose. Thus, it is important to describe the way such model, or the AI algorithm, contributes to the technical character of the invention. The same disclosure may be relevant to the assessment of inventive step as explained above.
Also there is a need for consideration under the doctrine of plausibility which is essentially a factual assessment of the assertion that one has achieved a claimed technical effect. This is conducted based on the level of disclosure and reproducibility thereof by a person of average skill and knowledge in India. To avoid such objections, one must provide the "best method" and clear evidence by way of examples which establish the achievement of the claimed technical effect.
CONCLUSION
With a large addressable market, relatively young population, large number of computer scientists and mathematicians, and world class ability in software, the future of research, development and consumption of AI clearly belongs in India. It is therefore recommended that authorities consider the foregoing and other challenges. Meanwhile, stakeholders, like us, will continue to do our bit by voicing concerns at different fora. All to roll out the red carpet for applicants and set the stage for India to become a leader in patenting of AI related inventions.
Footnotes
1. Ferid Allani V. Union of India, W.P.(C) 7/2014 & Cm Appl. 40736/2019, Decision Dated 12th December, 2019
2. V.B. Mohammed Ibrahim vs Alfred Schafraneck And Ors. on 4 June, 1958; AIR 1960
3. Shining Industries And Anr. vs Shri Krishna Industries on 3 May, 1974, AIR 1975 All 231
4. National Institute of Virology vs Mrs. Vandana S. Bhide, Pre-grant Opposition before the Controller of Patents in the matter of Patent Application 581 /BOM/ 1999
5. F. Hoffmann-La Roche Ltd vs Cipla Ltd., Mumbai Central, . CS (OS) No.89/2008 and C.C. 52/2008, Decided on 7 September, 2012;
6. Biswanath Prasad Radhey Shyam vs Hindustan Metal Industries Ltd (AIR 1982 SC 1444);
7. Press Metal Corporation limited vs Noshir Sorabji Pochkhanawalla, 1982 PTC 259(Bom);
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
