In keeping with a broad global trend, the Indian Patent Office has refused Dr. Stephen Thaler’s patent application which sought to recognise Dr. Thaler’s AI system DABUS as the inventor of a ‘food container and devices and methods for attracting enhanced attention’. The Indian patent office’s decision goes a step beyond the refusals issued by other jurisdictions and discusses, along with inventorship, the question of patentability of the claimed invention.
The refusal of the Indian Patent Office is a part of, and mirrors, a long line of similar orders issued by patent offices in 17 other jurisdictions. These applications have been filed as part of The Artificial Inventor Project. This is a pro-bono effort of lawyers from across the world who file, prosecute and litigate copyright and patent applications that name DABUS as the author/inventor as ‘legal test cases’ to challenge existing norms vis-à-vis protectability of AI-generated content. In this post I discuss the Indian Patent Office’s decision, possible next steps for Dr. Thaler in India and situate this refusal within the broader context of the scholarly challenge that AI-related patent applications raise for notions of inventorship and patentability.
DABUS Around the World
Device for the Autonomous Bootstrapping of Unified Sentience, or DABUS, is an AI system developed by Stephen Thaler. Dr. Thaler credits DABUS for autonomously generating an invention of a food container with inner and outer fractal profiles. It is claimed that this facilitates ease in holding the containers and allows multiple such containers to be stacked together for easier transportation. A single PCT application was filed in 2019 following from which applications were filed in a total of 18 jurisdictions. In these patent applications, Dr. Thaler lists DABUS as the sole inventor and himself as the applicant on behalf of the AI system. With this, he intends to challenge the human-centric requirement of inventorship.
The applications are usually refused. In the USA, UK, Australia, Germany and Switzerland they have been subjected to litigation, with varied results. South Africa has been the only jurisdiction to grant a patent without amendment in the application. The South African patent law does not require a substantive patent examination. In Germany and Switzerland, patents were granted on amendment of the named inventor from DABUS to Dr. Thaler. He submitted three auxiliary requests: first, a declaration that no designation of inventor was required; second, naming Dr. Thaler as the inventor with the description stating that the invention was created by DABUS; and third, that Dr. Thaler prompted DABUS to generate the invention. The third request was accepted in Germany and Switzerland with the effect that a natural person is required to be named as the inventor, and the fact that an AI system has been used to discover a technical teaching is acceptable. This does not have the effect of considering DABUS as the co-inventor but rather a tool to find the claimed technical teaching. The Swiss court also held that in training the AI system, providing input data and receiving the output, Thaler had meaningfully contributed to the inventive process, hence he could be named as an inventor. The threshold of human involvement required is, at present, an open question.
In Australia, the primary judge decided that an AI system or device could be recognised as inventor. The Full Federal Court reversed this opinion on appeal. This court held that the principle of accession or first possession, which broadly permits that the owner of a thing (tree) is also the owner of the benefits of the thing (fruits of the tree), does not apply. It was Dr. Thaler’s argument that since he was responsible for making, using and maintaining DABUS, he can derive title to DABUS’ inventions. At the European Patent Office (EPO), an addendum of the nature of the third auxiliary request was rejected. The Examining Division was of the opinion that the statement of this kind gave way to the implication of DABUS itself being the inventor. In the UK, after the UK Supreme Court found against listing the AI system as the inventor. Dr. Thaler filed another Form 7 in respect of a divisional application renaming himself as the inventor while continuing to acknowledge that DABUS conceived the idea of the fractal container. The UKIPO could not accept this ‘change of mind’. In appeal, the High Court of England and Wales agreed. Even though the IP office is generally not required to undertake any extensive inquiry on who is named the inventor, in this particular case, Dr. Thaler’s own admittance of his belief on the point of DABUS’s inventor status could be understood to mean that his turn-around of naming himself was “obviously defective”. The USPTO had similarly held, and the Court of Appeals agreed, that a machine cannot be named the inventor for not being an “individual” in the sense of a “natural” person.
Indian Patent Office Decision
The Indian patent office found lack of inventive step under Section 2(1)(ja). On the question of inventorship, the objection that an AI-based non-human entity cannot be recognised as a true and first inventor under Sections 2, 6, 7 and 10 of the Patents Act, 1970 was accepted.
On inventive step, the examiner found that prior art referring to containers with peripheral features like complimentary concave/convex surfaces, grooves, flanges, etc. was sufficient for a person skilled in the art (POSITA) to understand the claimed fractal geometry as a routine design consideration that could allow mechanical coupling and decoupling of multiple containers. The POSITA, likely a mechanical/packaging engineer or container-design technologist, would know about these routine structural variations, possible routine workshop modifications and use of materials to achieve these effects. Dr. Thaler also did not provide any experimental data or evidence to prove that the claimed fractal structure provided an unexpected technical advantage. The term ‘fractal’ was merely a descriptor of the container’s wall profile and in not providing any particular synergistic effect, it did not establish inventiveness. Independent Claim 1, thus, lacks inventive step and this hits at the dependent Claims 2-9.
On the question of naming DABUS as the “true and first inventor”, the Assistant Controller held that the existing patent law framework permits only natural persons to be recognised as inventors. Much emphasis was laid on the strict literal interpretation of the statutory connotation of “person” under the Indian Patents Act, 1970, the General Clauses Act, 1897, the Bharatiya Nyaya Sanhita, 2023, the analogous patent law provisions and judicial decisions on this point in the UK and the EPO. Cumulatively, the following are the requirements for being a natural person fit to be considered an inventor – a natural person capable of possessing legal rights, assigning such rights, and being represented after death; proof of right to make the patent application necessitates particulars such as nationality, address, and signature or a valid executed assignment; a “government” is a juristic ‘person’ when expressly statutorily identified as such; finally, legal personhood accompanies responsibility, intentions and accountability that are absent in machines. The requirement of proof of right derived from the inventor under Section 7 could not be justified by mere ownership of the machine. Here, the Indian office accepts the rationale of the Australian court in rejecting the accession argument. Reliance upon PCT Rule 4.17(ii) does not satisfy the statutory requirements under Indian law. A formal proof of right submission is mandatory for PCT applications under Section 138(4) of the Patents Act, 1970.
For a legal test case that has been solely aimed at pointing attention at, and generating dialogue on, human-centric inventorship norms and their black letter interpretation, the Indian Patent Office has comparatively gone a step beyond by conducting a more thorough investigation of the substance of the patent claims. On the requirements for inventorship, the decision aligns with the refusals issued in other jurisdictions.
On the Future of Indian AI-Patent Law and Policy
Two points in this document stand out for me which I think require more and better domestic discourse – first, the Assistant Controller’s analogy of a human-centric POSITA requiring a complimentary human-centric inventor; and second, the rather baseless and unsubstantiated policy decisions on AI & IPR advised in Report No. 161 of the Rajya Sabha, 2021 regarding the ‘Review of the Intellectual Property Rights Regime in India.’
The Indian office diverges in opinion on leaving some scope open for AI-assisted inventions. In the refusal, the Assistant Controller notes, “since inventive step is considered from point of view of a POSITA, the conception of invention should also originate from a human mind.” This could be at odds with the evolving state-of-the-art of AI systems and industry practices of using these AI systems as tools to generate patentable inventions. For instance, in the pharmaceutical sector, the use of Google’s AlphaFold Server has been documented in successful patent applications. The IP Office’s opinion in rejecting the DABUS application on this particular can potentially foreclose the space for accepting that some POSITA would understand how generative AI models work and are being used in developing new inventions. Like Germany and Switzerland, leaving a potential window open for acceptance of AI-assisted inventions and thereby starting a domestic conversation on the threshold of human intervention required, I think, would have made for a more dynamic decision; especially when questions of AI-related patentability are of direct importance to economic market dynamics.
Finally, a word on the 161st Rajya Sabha committee report. This Committee recommended “…that a separate category of rights for AI and AI related inventions and solutions should be created for their protection as IPRs.” Specifically, it recommended incorporation of AI-related provisions in the Patents Act, 1970 and Copyright Act, 1957 since these statutes “are not well equipped to facilitate inventorship, authorship and ownership by Artificial Intelligence”. An attempt to do this for copyright law was the DPIIT Working Paper, which suffered the brunt of pointed criticism. This Committee found requirement of human inventorship problematic – an argument that Dr. Thaler tried to use. The Assistant Controller was, of course, correct in separating policy from statutory law. In the 169th Rajya Sabha Report, which contains the actions taken by the government on recommendations of the 161st Report, at least the question of copyright authorship receives a more nuanced view. On AI overlap with patent law, the language is remains vague and uncertain. Domestic IP policy continues to remain of rather little help. In the meantime, it remains to be seen if Dr. Thaler fights this refusal in a court of law, as has been the modus operandi of the Artificial Inventor Project.