SPR3: AI Patents and the Abyss of Ambiguity

Despite a funding winter, AI companies have been raising money thanks to the explosive growth in this sector. With money swirling around, it is no surprise that the number of AI related patent applications at the Indian Patent Office (IPO) have also been rising. This is despite Section 3(k) of the Indian Patents Act stating that, “mathematics, business methods, computer programs per se, and algorithms,” are not patentable subject matter. Note that the term, “per se” has been attached only to computer programs and not to algorithms. Since most AI patent applications are essentially algorithm patents, they should be automatically disqualified. 

In this third edition of our Software Patents Review (SPR3), we take a look at a few AI related patent applications. In subsequent SPRs, we will also examine AI patents that have been granted by the IPO. 

Take the case of a patent application No: 202441025040 to implement a digital marketing strategy using artificial intelligence. Here the content for the marketing is created using AI applications, hence automating the process of content creation and marketing.

Pic 1: Field of Invention and Background information about the patent application no: 202441025040

Pic 2: Claims made in the patent application patent application No: 202441025040

This application was filed by an American company for their AI generative software which creates automated market strategies. Before we delve into the specifics of this application, it is worth noting that 63% of the granted patents belong to MNCs. It is hardly surprising that one of the applicants is the global head of one such MNC. In parliamentary discussions around Section 3(k), one of the big reasons for overthrowing the proposal that would have allied technical effects of software to be granted patents was the fear that most of these would go to MNCs. In many ways, MNCs taking the bulk of patent grants is an example of the gap between intent and reality when it comes to policy making in India. 

AI is set to be a horizontal technology that will be used by every industry. To allow individuals and corporations a 20 year monopoly on broadly used technologies will simply slow down the progress of our society. The irony is that most of the foundational technologies and software used in AI are Free and Open Source Software (FOSS), which is diametrically opposite to software patents. While FOSS gives users the freedom to share, modify and redistribute the software, patents are 20 year monopolies that give its owners the right to exclude others from using the techniques contained in the patent. In this specific instance, the patent seems to be merely a description of existing marketing methodologies with the word “AI” sprinkled on top. It is worth noting that, in the case of Yahoo v. Controller of Patents, the Indian Patent Appellate Board (IPAB) dismissed Yahoo’s claims on the grounds that, “Even the technical advance that is claimed over the existing art is only an improvement in the method of doing business.” 

A cursory review of other applications also shows a similar trend. See, for example, patent application number 202441023042

Pic 3: Claims made in Patent Application Number 202441023042

Again, this is simply sprinkling the word “AI” on top of a routine method of analyzing the process of legal outcomes. The three-step analysis of patent eligibility also leads us to a similar conclusion that this application is ineligible for a patent. 

The word patent originates from the Latin word, patere, which means "to lay open" or make available for public inspection. The term is a shortened version of “letters patent,” which was an open document or instrument issued by a monarch or government granting exclusive rights to a person, predating the modern patent system. During the industrial age, many inventors took their inventions to the grave. The net result was that others were unable to operate those machines. Therefore governments stepped up and offered a limited term monopoly (usually 20 years) to the inventor in return for disclosure of the workings of that machine. The inventors also had to deposit a working model of the machine at the patent office. 

The social contract between the inventor and the state was that inventors disclose their inventions, in return for a state granted monopoly for a limited period of time. After this limited period expires, the invention passes into the commons, where everyone can benefit from it. This contract is now broken in the case of AI, emerging technologies and software. 

Let’s go back to the AI patents as an example and see if the disclosure has any value. In the case of business method patents masquerading as AI patents, these are mainly existing processes on which exclusive rights are being sought. A functioning AI system needs software, training data and model weights to work. Patent disclosure might capture the business methods used, but the training data and model weights are usually left out. Therefore, anyone looking at these patents will not be able to replicate the patented ideas. In other words, the original social contact is broken. 

Granting a 20 year monopoly for such ideas is a bad bargain for society because these disclosures have no value. Anyone looking at these patents will not be able to replicate the same. The Indian Patent Office has not released any rules for reviewing AI inventions, unlike the patent offices in the US, UK, Singapore, Japan, and Europe. As a result, examiners are frequently uncertain about the methodology that must be utilized. They largely rely on subject matter exclusions relating to mathematical techniques/algorithms, business procedures, and computer programs per se when discussing AI techniques and functional applications, but they are not quite clear on how to apply such exceptions to AI inventions. Standards for determining the creative step and specification are inconsistent. Of course, this abyss of ambiguity is exploited by applicants. It is high time that the IPO updates the Computer Related Inventions (CRI) guidelines and takes a clear stand that coheres with our policy goals. 

Gagan and Athul Mohan are students of LLB (Hons), NLSIU.

Here are the links to the two Software Patent Reviews done by Gagan And Athul:

1. Systematic Method Of Analysing the process of legal outcome by using Artificial Intelligence
2. Artificial Intelligence-powered content creation in Automated Marketing Strategies