Patentability of Artificial Intelligence-Based Inventions in India
Introduction : AI has secured a significant niche in today's technology. It is utilized throughout numerous industries, for example healthcare, finance, transport, and web-based services. Today, a lot of new inventions are based on machine learning, data processing, and automated systems. Then comes the simple question of whether AI can bring about any patentable invention in India? On the other hand, AI is mostly algorithms and programs and the Patents Act, 1970, provides that a “computer program per se” cannot be patented. This becomes a bit complicated because AI predominantly uses software. However, AI is not always limited to software programs. It can also be found in machines and technical systems. That is, simply inquiring about whether the invention employs AI isn't enough. The issue here is whether the invention has a technical effect or it is only an abstract idea. This article attempts to provide a glimpse into the Indian approach towards AI-based inventions, especially with reference to Section 3(k) and why the concept of technical effect becomes important.
Nature of Artificial Intelligence Based Inventions
Typically, artificial intelligence implies machines that perform functions that require human reasoning. These are systems that learn from data and subsequently find patterns and produce output from the data. The vast proportion of these systems employ machine learning models and algorithms. This has brought revolutionary innovative technologies such as software inventions into the world of inventions driven by artificial intelligence. Many of them are based on data and mathematical models. This is problematic, because it is not possible to patent pure software and algorithms. But in reality, AI is hardly ever used on its own. It is used together with other systems. Doctors, for example, use AI to diagnose illnesses in medical equipment, while drivers implement AI to steer a car more efficiently, and manufacturers use AI to optimize their processes. Here, the invention is not only a computer program, but also a tech-based system. One then has to review the entire invention, rather than the software.
Position under Section 3(k)
The Patents Act, 1970, in its Section 3(k) provides that mathematical methods, business methods, computer programs per se and algorithms are not inventions. Obviously, the patent-excluding principle of software and abstract ideas can be seen in this case. At best, an AI invention will be rejected if it is an algorithm or a model only. Just creating a new machine learning model is not enough to get a patent. But the words “per se” are very important. They imply that only computer programs in their pure form are excluded. If this isn't enough, the invention can still be considered. For AI, this implies if the invention has a certain sort of actual technical advantage, this will be explored in greater detail. When, for example, a machine is improved by an AI system or AI makes a system faster or makes something better in some other manner, then it could not be considered just software.
Thus, patent law has not completely excluded AI patents, but it's not just software.
Importance of Technical Effect
It is crucial for the system to demonstrate a technical impact. Though it is not specifically designated by the law, it can be applied in practice to determine whether something is patentable or not. Technical effect basically refers to the invention being useful technically. It needs to enhance a system or process or solve a problem. This can occur in AI inventions by making a machine better operate, faster, or more efficient. For example, if an AI recommendation to make a machine more efficient or less prone to error is implemented, then it can be considered a technical effect.
However, if the disclosure of the invention does not bring the information and/or the results to technical use, the invention may be refused. If there is no clear definition, this will vary in relation to the invention's explanation.
Approach of the Indian Patent Office
The Indian Patent Office has laid down some guidelines for computer related inventions. The latter are referred to as CRI Guidelines. This is the rule as per these guidelines, if the invention is just a computer program or an algorithm, then it will be rejected. But if it shows practical application or related to a piece of hardware equipment it may be further investigated. In the case of an "AI invention" this also implies that if a stakeholder is claiming just a model or algorithm, it will be rejected. If the invention demonstrates actual usage of AI in a system, then it's worth considering. But the Patent Office generally will concentrate on the substance of the invention, not the words. However, making simple changes in the application will not assist. The contribution has to be a technical one.
For this reason, the question becomes extremely significant.
Judicial View
In India there are only a handful of cases that are specifically concerning AI patents. However, courts have given clarity on the interpretation of section 3(k). Passage of the Delhi High Court in the case of Ferid Allani v Union of India, 2019, is of significance.
The Court in the Farid Allani case however said that mere inclusion of software should not be a bar to invention. This shows that the wording alone is not decisive, the drafting of the patent application becomes important. This is advantageous in the context of AI since it demonstrates that inventions that rely on software should not be discredited. The Court stressed that it is the technical aspect of the invention which is relevant and not the presence of a computer program.
Practical Difficulties
There are on-going problems here despite guidelines and opinions from the courts that have provided guidance. There is a problem in sometimes being able to see the difference between an algorithm and a technical invention. The other issue is the lack of definition of technical effect. This may lead to the possibility of conflicts arising between examiners.
Another big factor for them to take into account is the fast-paced nature of AI. It's constantly developing as it gets additional training. When a patent is applied for, usually an invention is set out when the patent is applied for. However, the AI's performance can vary over time depending on updates and retraining of the algorithm. Thus, the object of protection becomes difficult to determine. Additionally, there is the ownership and contribution question. At times, the AI systems come up with solutions or outcomes without human intervention. This can produce dilemmas regarding the development, the user or the machine itself. In India, patent law is still based on the principle that an invention is a product of human activity. This can lead to some confusion in dealing with Ai outputs.
Another challenge is the data that is needed for AI creation of inventions. In some instances, an algorithm might not need to be improved, but instead it may be necessary to analyze the data. But just data is not patentable. Once again, it is hard to determine which of the inventions is novel and worthy of protection. All this makes AI patents not only a legal, but also a technical challenge. A clear demonstration of how the invention works, and its technical effect is required. A wealth of data is essential for the growth of AI. The new idea may not be clear, whether part of the algorithm or the data employed. In some cases, the developer may choose to find the information of the software and then claim that it is a hardware device. In some instances, the developer might attempt to sell the software as a hardware device to avoid rejection. This can create confusion. All of this makes patenting novel inventions of AI in India still difficult.
Conclusion
The interpretation and application of Section 3(k) is the primary determinant in Patentability of Ai inventions in India. As per the law, computer programs by themselves are not patentable. Because the computer program is itself unlike patentable, there is no patent protection for pure AI algorithms. Meanwhile, patents for AI are not totally prevented by the law. If the invention falls under the tech related (as it makes a tech impact or provides a tech solution) then it's acceptable. In both cases, the Indian Patent Office and the Ferid Allani case, these have reiterated that the emphasis needs to be on the tech aspect of the invention. With the emergence of AI, this will be more applicable. We need innovation while ensuring open access to basic ideas and the law will need to strike the balance.
Author :- Adarsh Yadav, in case of any query, contact us at Global Patent Filing or write back us via email at support@globalpatentfiling.com.
References
- The Patents Act, 1970, § 3(k), Government of India Available at: https://ipindia.gov.in/writereaddata/Portal/ev/sections-index.html
- Ferid Allani v. Union of India, Delhi High Court, 2019
- Guidelines for Examination of Computer Related Inventions (CRI), Office of the Controller General of Patents, Designs & Trademarks
- Office of the Controller General of Patents, Designs & Trademarks, Manual of Patent Office Practice and Procedure
- World Intellectual Property Organization (WIPO), Artificial Intelligence and Intellectual Property Policy Available at: https://www.wipo.int/about-ip/en/artificial_intelligence/