Author: Harsh Kashyap, School of Law, Forensic Justice and Policy Studies, National Forensic Sciences University, Gandhinagar
To the Point
Advancements in Artificial Intelligence (AI) have surpassed a mere data analysis or automation tool. Today’s AI systems, through machine learning and neural networks, can generate inventions from new pharmaceutical products to engineering solutions largely free from human input. This development fundamentally challenges philosophy and law around patents.
The critical question for us is: Could AI be understood to be an inventor in Indian Intellectual Property (IP) law (specifically the Patents Act, 1970)? While the legislation assumes that the inventor is a human being, technology indicates otherwise. AI systems are capable of independent creativity. The global case law surrounding the DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) cases illustrates the issue. India has yet to confront this issue directly, but soon, with the growth of our AI ecosystem, it will no longer be able to take a silent position on this question.
Abstract
The issue of whether AI should be considered an inventor has become an increasingly important debate in the law of patents. Although inventorship is limited to natural persons under India’s Patents Act, 1970 and similar regimes around the world, it is acknowledged that this limitation could stem from a belief that only natural persons can possess “intellectual creativity.” The DABUS litigation and challenges to traditional notions concerning inventorship raised by AI-generated inventions challenge this perspective. This article explores Indian patent law, relevant case law and highlights how current legal regimes may not allow for AI to be an inventor yet they could accommodate a legislative amendment to not only preserve the concept of natural persons inventing but also potentially formally recognize and accommodate the input and contributions of AI.
Use of Legal Jargon
AI created inventions challenge some of the doctrinal assumptions which underlie patent law. The Patents Act, 1970 is based on human understandings of inventorship, ownership and accountability. The Act assumes human actors exist by virtue of Section 2(1)(p): “Patentee” means a person to whom a patent is granted; Section 6(s) stipulates the application must be made by the true and first inventor, or the assignee of the inventor, and Section 10 lists how the patent specification must disclose the invention on which the patent is requested and to be signed by the inventor. All of these require a natural person who has the legal capacity to act.
AI based inventorship creates tensions with doctrinal assumptions and associated requirements under patent law such as:
Inventive Step – will AI output satisfy the standard of being non-obvious, where machines analyze data sets a human cannot possibly process?
Disclosure Requirement – AI cannot make the requisite statutory declaration under Section 10 of the Act, and the inventor’s signature is clearly part of the obligation.
Existence of Ownership – AI has no legal personality, therefore cannot assign rights, raising ownership issues.
Principle of Accountability – patent law operates on the assumption some wrongful acts of the inventor must be accounted for through the patent system, e.g. misrepresentation, fraud etc. AI lacks the mens rea.
Therefore, AI inventions challenge the person-centric foundations of patent law.
The Proof
Current Statutory Framework of India
Although the Patents Act, 1970 makes no mention of AI being acknowledged as an inventor, it can still be read to exclude AI:
Section 6: The sole applicant can be the true and first inventor or assignee. AI is not a legal applicant.
Section 10: The complete specification requires a signature. AI cannot sign.
Section 2(1)(p): “patentee” means person, with the term person reinforcing there must be a natural person.
Exclusions whose risks include
Innovation gap: Denying AI inventorship may diminish investment in AI leveraged research
Uncertainty in ownership: With no clear rules, conflicts will inadvertently arise among developers, users and corporations.
Delay in dealt with doctrinal change: Patent law will become unusable with the courts unable to grapple with disruptive technologies.
Arguments against AI as an inventor
No legal personality: AI cannot own, license or enforce any rights
Patent theory: Incentives are based on the theory of IP law being to encourage human registrants, AI cannot receive an incentive.
No accountability: Patent priority fraud or infringement needs to be able to identify an entity that can be legally accountable.
As it sits, Indian law does not grant inventorship to AI. But the realities of industry demand a check to the current paradigm.
Case Laws
Thaler v. Comptroller General of Patents (UK Supreme Court, 2023)
Facts: Dr. Stephen Thaler applied for patents with the UKIPO, giving DABUS, an AI, consecutively as inventor. The UKIPO rejected Thaler’s application.
Held: The Court interpreted the “inventor” as a natural person only under the UK Patents Act 1977.
Relevance: The decision confirms strict statutory interpretation – no AI could ever be inventor.
Thaler v. Vidal (U.S. Court of Appeals, 2022)
Facts: Thaler made equivalent applications in the U.S. with DABUS as the inventor. United state patent and trademark office (USPTO) denied Thaler’s application.
Held: The Court found that according to U.S. law, only a natural person can hold inventorship in the U.S.
Relevance: Provides further support to strict human-centric forms of inventorship as defined in the various U.S. statutes.
Thaler v. Commissioner of Patents (Federal Court of Australia, 2021/2022)
Facts: The initial Australian Federal Court ruling had allowed an AI to be accepted as an inventor, but this was overturned by the Full Federal Court, holding that a final decision could only exist for humans.
Relevance: The brief judicial consideration of the possibility of an AI as inventor returned to an anthropocentric viewpoint.
DABUS Patent in South Africa (2021)
Facts: The patent office of South Africa granted a patent naming DABUS as inventor – without an examination.
Held: The patent was granted procedurally not judicially.
Relevance: Only country that has formally accepted AI inventorship, there has been no scrutiny.
Ferid Allani v. Union of India (Delhi High Court, 2019)
Facts: Related to the patentability of computer-related inventions rejected by IPO.
Held: Court pointed to the interpretation of patent law to promote technological innovation.
Relevance: Not directly about AI inventorship, it’s an indicator of openness of Indian courts to adapt to emerging technologies.
Comparative Perspective
United States & United Kingdom: Direct human inventorship is require.
European Patent Office: Takes the same position as the UK, requires inventorship by natural persons.
Australia: Had a short dalliance with AI inventorship, which was later overturned.
South Africa: Outlier, and accepts AI inventorship without examination.
India, as an emerging technology hub, must decide whether to take a conservative approach or set up a flexible platform for AI inventorship.
Conclusion
The AI inventorship issue exists at the crossroads of law, technology and philosophy. AI clearly generates inventions that meet the tests of novelty, inventive step and utility, but, under the Patents Act, 1970, inventorship is limited to natural persons and AI outputs will never receive formal recognition.
Denying formal recognition to AI as an inventor will not only limit innovation but, if granted, raises accountability and ownership issues. A possible pragmatic solution is to keep human inventorship in the public record (rights vests in the developer, deployer, or assignee), but acknowledge AI’s role, with inventorship retaining its traditional view. At some stage, India could adopt sui generis legislation for AI-generated inventions, providing doctrinal clarity and a means of guaranteeing global competitiveness.
So, in terms of current legal standards, AI cannot be credited as an inventor in India. But legal reform seems unavoidable, and the sooner that India envisages this it will certainly put itself in a better position in the global IP eco-system.
FAQS
1. Can AI be an inventor according to Indian law?
No. The Patents Act, 1970 requires there to be a natural person included in the definition of invention as inventor. When we mention a “true and first inventor” we are making reference to a human being.
Q2. What if the invention is made by AI without assistance from a human?
AI generated inventions can be patented but only if a human accepts recognition for themselves as the “inventor” of the AI (the owner or developer of the AI).
Q3. Which countries have accepted AI as an inventor?
Only South Africa has accepted it and not without hesitation. The U.S. (as well as the EU), UK and Australia have all rejected AI inventorship.
Q4. What is the risk of rejecting AI inventorship?
It may stifle innovation altogether, create ownership and patent disputes, and ultimately take patent law further away from where it needs to be in step with technological advancement.
Q5. What sort of reform may be needed regarding patent law in India?
Reform may include changing the definition of invention, taking into account AI induced inventions or, recognizing AI product outputs as unique sui generis rights.
