Artificial Intelligence and Intellectual Property Rights: Navigating the New Legal Frontier

Abstract

Rapid advancement of Artificial Intelligence (AI) technologies have caused substantial changes in the field of Intellectual Property Rights (IPR), impacting on the basic legal principles including those of authorship, inventorship and ownership. Traditional IPR systems, such as copyright, patent, and trademark laws, have been designed under a paradigm that assumes a human approach to creativity and ingenuity. With the advent of AI systems that can create literary compositions, music, art, and even patentable products with very little or no human involvement, however, these assumptions become legally problematic. This paper explores the relationship between AI and IPR, highlighting the shortcomings of the current legal framework in India and in some international jurisdictions to accommodate AI-produced outputs as

protectable IP. The paper identifies critical gaps in the current IPR legal landscape by comparing and contrasting judicial statements, laws and policies from the U.S., U.K., EU and India, and recommends a multi-faceted normative approach to fill these gaps. Without specific legislative changes, sui generis IP protection and internationally agreed-upon standards, the benefits of AI innovation will not be sufficiently encouraged, legally defined and subject to relevant laws and

precedents. In the Fourth Industrial Revolution, it is contended that a rethinking of authorship and inventorship doctrines, together with the introduction of a limited protection regime for AIs, is both needed and judicious.

Keywords: Artificial Intelligence, Intellectual Property Rights, Copyright, Patent Law, Authorship, AI-Generated Works, Sui Generis Protection

I. Introduction

With the dawn of Artificial Intelligence, there was a shift of paradigm across almost all sectors of human endeavours. Today, AI systems are doing jobs that were once uniquely human, such as autonomous driving or diagnosing patients. This change is more jurisprudentially difficult

anywhere than in the field of Intellectual Property Rights. 1 It is the anthropocentric basis on which the present IP framework was built: only human beings are entitled to legal protection for their intellectual products as authors, inventors or creators.

This is the current premise that is being attacked. Artificial intelligence systems like DALL-E, GPT-4, AIVA and IBM's Watson have shown the ability to write novels, compose symphonies, create artwork and develop pharmaceuticals — and all of these would clearly be eligible for IP protection if done by a human. But the law hasn't caught up with this technological fact. The legal world around the globe is struggling to answer the question of whether a computer can be an author or an inventor and, if it cannot, who should be the ones to own the rights to a work created by an AI?

This paper has six parts. Part II gives an overview of the state-of-the-art IPR framework and its anthropocentric basis. Part III offers a comparative law analysis of the legal responses from key jurisdictions to AI creations and inventions. Part IV discusses the Indian Law System and the state of the Indian IP statutes. Part V discusses the key jurisprudential challenges and proposes normative reforms. Part VI ends with recommendations for a future legal framework.

II. Background: The Anthropocentric Foundations of IPR

The Anthropocentric Foundations of IPR consists of three parts. Three-parts make up the Anthropocentric Foundations of IPR. At heart, IP law is a system of legal incentives that creates a system of exclusive rights to the output of human creativity and innovation for a period of time. Each of the three major IPRs – copyright, patents, and trademarks – makes an implicit assumption of human agency.

A. Copyright Law

Original literary, artistic, musical and dramatic works are covered by copyright. Throughout copyright history, “originality” was defined as the need for a human author who put some original skill into the creation. In the United States, the Supreme Court ruled in Feist Publications, Inc. v. Rural Telephone Service Co. that originality requires “at least some minimal degree of creativity” that is "emanating from the author. In India, the Supreme Court in Eastern Book Company vs. D.B. Modak held that the standard for creativity of the authors was to be the modicum of creativity, which means the intellectual effort of the author.

The problem with works generated by AI is that they can be original in a pragmatic sense, they can be new and complex, they can be even aesthetically pleasing, but there's no meaningful way to link them back to the human author's creative decisions. The AI system produces outputs by making inferences based on the training data, rather than by consciously and creatively deciding about what to produce. The originality is functional, without the stamp of the author, and leaves a doctrinal gap.

B. Patent Law

Under patent law, an inventor has exclusive rights in an invention that is novel, non-obvious and industrially applicable. It is the generally accepted definition that the term “inventor” refers to a natural person. In the case of Thaler v. Vidal, the United States Court of Appeals for the Federal Circuit ruled that the inventor must be a natural person under 35 U.S.C., and that an AI system named DABUS cannot be named as the inventor of a patent. The UK Supreme Court reached the same conclusion in Thaler v. Comptroller-General.

Stephen Thaler created an AI system called DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) that created two inventions by itself: one for a food container with a fractal surface to improve heat transfer in thermoclines and a second to create a flashing attention- seeking light in emergencies. The United States, the United Kingdom, the European Patent Office and India's patent office deemed the inventor's name, DABUS, not to be a human being

and rejected these applications on that basis.

C. Trademark Law

Trademark law provides exclusive rights to a person who uses a distinctive sign in commerce to indicate a source of his or her goods or services. The Indian Trade Marks Act, 1999, like its American and UK counterparts, (the Lanham Act, 15 U.S.C. 67; 1051 et seq. and the Trade Marks Act, 1994, respectively), tacitly assumes a natural person or an incorporated legal body as the applicant and registered trademark owner. With AI systems becoming more prevalent in creating trademarks, including brand names, logos, slogans, and product packaging (what may be called; AI-generated trade dress), a new set of questions arises about the ability to use, register, and enforce trademarks in AI-generated trademarks. 11

The focus of the distinctiveness under trademark law is the ability of a sign to direct the mind of the average consumer to the commercial source of products or services. The distinctiveness analysis is not based on human creativity, but the trademark applicant needs to be a natural

person or legal entity that can carry out business and initiate infringement actions. An AI system is not a legal person, and thus cannot use a mark in commerce in the legally cognisable sense, nor can it maintain litigation. The key issue will be whether the human user of a mark created by an AI, which is copied from the AI's own generation and then incorporated into their own trading, becomes the sole proprietor of that mark under the current law. The answer is yes if the human was involved in selection and adoption, but this does result in a de facto human intermediary between the generative process of the AI and the trademark. This intermediate step creates more questions regarding real usage of a trademark when the trademark is not created by a human and when the human's work might consist only of copying and placing the output of the AI into commerce.

D. Trade Secrets and Confidential Information

A trade secret is a type of IP protection that does not require registration and does not have a specific term, but is limited to the confidential nature of the information and reasonable steps taken to maintain its confidentiality. Unlike the United States, which adopted the Defend Trade Secrets Act in 2016, India has no dedicated statute on trade secrets and instead relies on elements of common law principles, such as breach of confidence, and contractual obligations of maintaining confidentiality. The Information Technology Act, 2000 gives minor ancillary protection to unauthorized access to digital systems containing trade secret information.

The architecture of an AI system, the training data used, the weights of the model, and hyperparameters are often one of the technology enterprise's most valuable trade secrets. A large language model trained on proprietary corpora can be an extraordinary significant trade secret. A

regulatory versus proprietary secret clash is a critical but largely unexamined legal battle between obligations to be transparent about AI systems imposed by regulatory regimes, such as the EU's AI Act (2024) and India's upcoming Digital India Act, and to safeguard confidential information that can be protected as a trade secret. Courts will be asked to consider the exact limits at which the trade secret protections of the AI developer's algorithm disappear once the developer has disclosed its functioning in order to comply with regulatory requirements or for IP registration purposes, and to decide whether disclosure of some of the algorithm's functioning is enough to meet the statutory requirements without losing the trade secret protection.

III. Comparative Legal Analysis

A. United States

According to the U.S. Copyright Office, the United States Copyright Act gives copyright in authors which excludes non-human entities. In Naruto v. Slater, the Ninth Circuit reasoned that animals do not have standing to sue under copyright law, which has been analogized to AI systems. 15 In a recent update of guidance on AI-generated works, the U.S. Copyright Office states that works created without a human author are not subject to copyright protection, but works with sufficient human authorship may be.

On the patent front, the Federal Circuit's decision in Thaler v. Vidal precludes patentability of AI inventions under the existing law. Some scholars have called for a machine creativity doctrine, which would grant authorship or inventorship to the human who operates or deployed the AI system, and for legislative changes to incorporate that concept. 17 The Google LLC v. Oracle America decision, which broadened the fair use doctrine in the software arena, added to the confusion and had implications for AI training on copyrighted data sets.

B. United Kingdom

The UK Copyright, Designs and Patents Act 1988 (CDPA) is significant because of its specific wording in this area which grants copyright protections to "computer-generated works (Section 9(3)) to the "person who undertakes the necessary arrangements for the creation of the work. This is an exception that is not seen in other major IP jurisdictions, and it makes it possible for someone to be the copyright holder of an AI-generated piece of work, as the AI-generated work would be credited to that individual who made the AI-generated output possible.

In the patent arena, however, the UK Supreme Court in Thaler v. Comptroller-General definitively rules out AI inventorship: Only human can be the inventor under the UK Patents Act 1977. This imparts a curious imbalance to UK law: AI-produced artistic works can receive copyright protection while AI-produced inventions can receive patent protection only if a qualifying human inventor can be found.

C. European Union

The European Parliament decided to investigate how AI-generated works can be protected, and for whom, in its Resolution of October 2020 on IP Rights for the Development of AI Technologies. 20 The EU has also been cautious, stressing that a new protection regime should not stifle human creativity or lead to perverse incentives. The European Patent Office has all along refused to recognize AI as inventorship, following the reasoning of the Federal Circuit. Although the EU Artificial Intelligence Act (2024) is mainly a regulatory tool, it also has IP consequences, such as the requirements for transparency and data governance concerning the training of AI systems, which is linked to text and data mining and copyright law.

D. India

The Indian IP regime includes Copyright Act, 1957, Patents Act, 1970 and Trade Marks Act, 1999 etc. None of these laws includes AI-generated work or AI inventorship. The Copyright Act's definition of author (Section 2(d) applies to certain categories of human creators; although Section 2(d)(vi) lists computer" as a category of author for computer-generated works, it is interpreted to mean “computer operated by human” and not “computer operated by AI”.

Although ambitious, India's National IPR Policy (2016) does not cover the creation of IP by AI. This lack of legislation or judicial precedent also leaves India with a lot of legal uncertainty for AI developers and deployers, especially with the country's growing AI startup scene and its goal of becoming a global tech hub.

E. China: A Pragmatic Comparative Perspective

The Chinese jurisdiction has become one of the most significant in the AI and IPR debate, as China's domestic AI sector continues to rapidly expand, and its stance on IP law is particularly pragmatic. In Tencent Technology Co. Ltd. v. Shanghai Yingxun Technology Co. Ltd. (2019), the Shenzhen Nanshan District People's Court determined that an article created by Tencent's Dream writer AI system could be protected by copyright as it was the result of the human operators creative arrangement and intellectual investment into the design and operation of the AI. The court ruled that Tencent is the author of the creative work, because it initiated the creative process, organised it, and technically implemented it.

While the 2020 amended Copyright Law of China does not specifically mention AI-generated works, historical copyright legislation and guidance from the National Copyright Administration of China (NCAI) suggest that the copyright of AI-generated content should be allocated to the legal person or natural person who arranges the creative work that underlies the operation of the AI. Within the patent arena, the China National Intellectual Property Administration (CNIPA) has affirmed that AI-assisted inventions must be disclosed in compliance with the standard, as well as the requirement that AI-assisted inventions must be made by a natural person, but has also called for further research into AI-assisted inventions and the disclosure requirements for inventions created by AI.

China's strategy which is doctrinally less rigid with a greater emphasis on operational considerations and legal flexibility, and more focused on the motivation of domestic AI leaders is instructive in contrast to the more doctrinally rigid positions of the Western jurisdictions. The Chinese approach is to credit AI-generated IP to the human or legal entity that created the AI's creative arrangements has been a practical compromise for emerging economies, such as India, where the AI sectors are in similar developmental phases. It will not be doctrinally rigid like US Federal Circuit's stance in Thaler v Vidal or have a regulatory framework of a sui generis regime, but will offer legal predictability to ensure commercial investment in AI use in creative industries.

F. Australia: Judicial Evolution and Legislative Consultation

Australia has been unusually bold at first instance but has been a retrenchant at the appellate level in its relationship with AI and IP. The Federal Court held in Thaler v Commissioner of Patents [2021] FCA 879 that an AI system could be an inventor under the Patents Act 1990 (Cth)

as the term inventor is an agent noun, and should be interpreted as it would be in natural language without being restricted to human entities. The court also held that it was essential to assign the AI as inventor to avoid creating a where AI created inventions would enter the public domain, thereby depressing the incentive to invest in AI-assisted inventions. The result of this decision was that it was the first time that a court of competent jurisdiction in a common law jurisdiction found an AI to be the inventor.

In Thaler v Commissioner of Patents [2022] FCAFC 62, the Full Federal Court unanimously disagreed, ruling that the concept of inventor in the Patents Act 1990 (Cth) must be a natural person, and that the Act's naming of inventors and assumption of applicant rights by inventors required a human element at every stage. Refused special leave to appeal by the High Court of Australia. In spite of this, the judicial dialogue stimulated by the Thaler case has stimulated debate on the policy agenda on how to respond to AI inventorship, with IP Australia undertaking formal consultations to consider whether there should be any changes to the legislation – meaning that Australia is now one of the few jurisdictions taking overt steps towards a “legislative response” to AI inventorship beyond mere judicial interpretation of existing laws.

IV. Key Jurisprudential Challenges

A. The Authorship and Inventorship Vacuum

The biggest problem is the authorship/inventorship problem: If an AI system creates a work or invention on its own, there is no obvious person in the traditional sense as an author or invento. As per the current law's unfair to claim the AI to be the author of the created content. It is factually difficult to state that the AI's output was caused by the programmer who created the AI if the output was not foreseeable or controllable by the programmer. It's difficult to credit the user who initiated the AI if they didn't do much. There needs to be a solution to this three-way uncertainty.

B. Training Data and Copyright Infringement

AI systems are trained on large datasets of copyrighted text, images and music, which are frequently collected without the consent of the copyright owners. It is still subject to an active lawsuit whether such training is infringing upon copyright or is covered by the exceptions to copyright, including the fair use doctrine in the U.S. and text and data mining in the EU. The text produced by these AI tools could also be plagiarized from the training materials, which also introduces additional infringement issues.

In the U.S., various authors and artists are currently taking legal action against AI companies, such as OpenAI, Stability AI and GitHub Copilot, in several class action lawsuits. These rulings will significantly shape the future of AI in India and beyond.

C. Ownership and Commercialisation

If a work produced by AI were considered "creative" and therefore eligible for protection, the issue of ownership of the resulting IP rights would still be unclear. There are several competing parties: the AI developer, the AI deployer (business that uses the AI), the user that gave the prompt that the AI will use. The incentive effects vary by ownership allocation, and in the absence of clarity in the law, may necessitate tailor-made contractual solutions.

D. International Harmonisation

IPR is territory-based, while AI is global based. The product protected in the USA, trained from the data of Europe and deployed in India is protected in India or not, or infringing in both countries simultaneously. The lack of a harmonised approach on AI and IPR leads to forum shopping, regulatory arbitrage and market fragmentation, which are not conducive to innovation or legal certainty.

E. Liability for AI-Generated Infringement

A relatively under-regulated area of the AI-IPR nexus that has seen much judicial activity is the issue of accountability in situations where AI systems produce outputs that violate the rights of third parties. The generative AI ecosystem typically includes three types of participants: the AI developer, who designs, trains and maintains the AI system; the AI deployer, who uses the AI system via an API or commercial platform, and incorporates the AI into a product or service; and the end user, who supplies the specific prompt or instruction that results in the allegedly infringing output. Under none of the major jurisdictions laws is it settled who is legally responsible for the infringing result when it is produced by the AI.

Direct infringement under Section 51 of the Copyright Act, 1957, which says that any person who does any act which the statute only provides for the copyright owner is an infringer, and secondary infringement may also be applicable under Indian copyright law. Drawing on previous English law, the authorisation liability doctrine has been recognised in Indian courts and could potentially hold AI developers and deployers liable for infringed upon work that they are “aware of”. The same kind of questions of contributory and vicarious infringement are raised in ongoing litigation against companies such as Stability AI, OpenAI and GitHub Copilot in the United States. The decisions of these proceedings, especially on the issues of substantial similarity in AI-generated work, if it was used to generate works through training data that can be used for direct infringement, and the extent of secondary liability for the developers, will inevitably create an analytical framework for Indian courts and legislators in cases of similar claims in the future.

This is further complicated due to the probabilistic and stochastic nature of the output of generative AI. An AI system does not make an output in the way that a human would copy, but rather uses statistical procedures and therefore should not be called a copy in the traditional sense of the word. The question whether the elements of infringement, namely fault and mens rea are met if the AI developer chooses to train the AI with copyrighted works, even without knowledge of reproduction in individual outputs, is one that is not satisfactorily answered by the existing doctrine and which must be answered by a legislative action with doctrinal precision.

F. Moral Rights and the Integrity of AI-Generated Works

Moral rights are a philosophically separate area of IPR doctrine that includes rights of attribution (paternity) and rights of integrity. The moral rights are thought to attach to the artist as a person, they cannot be transferred or commercially negotiated like are the economic rights. The Berne Convention for the Protection of Literary and Artistic Works (1996), to which India is a party, provides protection for the right of the author to claim authorship and to object to any changes made to their work that are harmful to his honour or reputation. These obligations are reflected in India's Copyright Act, 1957 – Section 57 author's special rights – right of paternity and the right of integrity which endure beyond the economic rights of the author and subsist even after assignation of copyright.

The issue of moral rights for AI-generated works poses a jurisprudential conundrum of the highest calibre. Moral rights are based on the assumption that there is a human creator whose reputation, artistic vision and personal integrity are embodied in the work and therefore worthy of protection by law as part of the personality of the creator. An AI system has no reputational standing, and cannot be dishonoured by misattribution, nor be harmed by the mutilation of its production. The key philosophical argument for moral rights – that the protection of the creator's persona is coterminous with the creation – is essentially detached from the legal form in which it exists.

On the other hand, if AI-generated works are considered to have no moral interests whatsoever, derivative works created based on the original works generated by AI can be modified, mutilated, miscredited or used commercially without any legal recourse, reducing the integrity

and quality of AI generated intellectual output in the marketplace. Legislative intervention, in the form of outright exclusion of moral rights from AI-generated works, and/or development of a restricted version of the integrity right in the hands of the deployer, is both overdue and feasible within the confines of the Indian copyright law.

G. AI Acceleration and the Problem of Patent Evergreening

The AI Acceleration and the Patent Evergreening Problem.The Problem of Patent Evergreening in the Era of AI Acceleration. Another systemic threat that comes from AI-assisted invention is the capacity of large pharmaceutical and biotechnology companies to use AI to create hundreds or even thousands of patented versions of existing drugs or biochemicals, extending the patent's lifespan well past its expiration. This practice, commonly referred to as “patent evergreening,” predates AI, but generative AI greatly speeds up the scope and sophistication of this practice. A conventional medicinal chemist could create 10 to 20 lead structure variations a year; an AI system could produce hundreds of thousands of new lead structure variations in days, with each option meeting the statutory requirements of novelty and non-obviousness.

With such an easy route to patenting pharmaceutical variants through AI, drug companies can systematically exploit AI to claim monopoly rights over entire drug classes, thus preventing generic competition and keeping prices high, particularly in markets where people are sensitive to costs like India. Section 3(d) of the Patents Act, 1970 of India already exists in the form of a provision which if read with the Supreme Court&'s judgment in Novartis AG v. Union of India (2013) provides protection to new forms of known substances only when there is “an adequate basis of distinction” between the new forms and the known ones. The effectiveness of Section 3(d) in combating AI-powered evergreening has not yet been tested by judges, however, and the patent offices will need significant additional technical capability to identify AI-schemes of evergreening for the entire range of structural chemistry and molecular biology.

V. Discussion and Proposed Normative Framework

The foregoing analysis shows that the in place IPR framework is structurally inadequate to regulate AI generated outputs. This calls for action at the legislative level, at the judicial level and at the international level. The following structure is suggested.

A. Legislative Reform: Sui Generis Protection

The most doctrinally pristine option is to develop a sui generis IP regime that is separate from copyright and patent laws, but draws on the principles of both. This would place a ‘substantial investment’ requirement on the human subject, be it a human individual, company or institution, to create the AI system and use it to produce the protected output, and grant a shorter term of protection for these outputs (e.g., 15 years for AI-generated works compared with the copyright term of life plus 60 years in India). It would be similar to EU's sui generis database right and would offer a viable, pragmatic remedy that doesn't affect current IP doctrine.

B. Expanded Statutory Definition of Authorship

Meanwhile, India needs to incorporate computer-generated works in the definition of author as per Section 9(3) of the UK CDPA, and copyright ownership should be accorded to the person who makes the necessary "arrangements" to create the AI-generated work. This would give legality without an overhaul to the law.

C. AI Disclosure Obligation

The disclosure of the use of any AI system to produce a work or invention should be mandatory for any IP registration system. This transparency requirement would enable IP offices, courts and

the public to evaluate the human contribution and the level it reached, and then adjust the level of protection.

D. International Treaty Framework

The negotiation of a new international treaty – similar to the Berne Convention or the TRIPS Agreement – exclusively concerning AI and IPR should take place within the WIPO framework. Major AI market and a signatory to the TRIPS Agreement and the Berne Convention, India

should be actively involved in the development of this new international framework. Such a treaty should cover the ownership vacuum, the training of data rights, the allocation of ownership and minimum standards of protection.

E. Fair Dealing for AI Training

The copyright law in India needs to be updated to incorporate a specific exception for the use of text and data mining for the purposes of training AI models, with appropriate conditions such as honouring the rights of the source, ensuring that it is not used for commercial gain or payment to a collective licensing fund that pays authors and publishers. This would strike a balance between allowing AI innovation and respecting the rights of human creators who generate the underlying

content of AI training sets.

F. Collective Licensing and Remuneration Schemes for AI Training Data

One of the most intriguing proposals for balancing the interests of human creators and AI developers is that of establishing a collective licensing system for AI training data. The concept of collective rights management, which has been successfully implemented in the music industry in India through societies like the Indian Performing Right Society (IPRS) and Phonographic Performance Limited (PPL) in India, and similar organizations in other countries, provides a tried and tested model for efficient licensing of huge amounts of works under copyright. With a collective licensing scheme of AI training data, AI developers would contribute levies to a centralized data pool, administered by a rights management body under statutory oversight, and the rights holders would be paid depending on the extent of the use of their works in AI training corpora that can be measured.

This could either be a statutory compulsory licence – which would require everyone who holds a copyright and all AI developers within the jurisdiction to participate, even if they don't consent or an extended collective licence, that would also apply to those who are not members of the rights management organisation, but once enough rights holders have joined, it would apply to everyone beyond that. In its consultations for 2022-2023 on AI and IP, the UK Intellectual

Property Office considered a similar collective licensing arrangement, but rejected it in its current form. In India, with its diverse creative economy, and where collective rights may not be as well established as in other jurisdictions, design of the legislation will have to ensure fair

compensation for a wide range of stakeholders, including authors writing in any of the constitutionally recognised languages, classical musicians, folk artists, regional cinema producers, and traditional knowledge holders whose works can be incorporated into AI training

datasets without their knowledge or consent.

G. The Role of the Judiciary and Common Law Development

That comprehensive legislative solution is the "best" long-term solution to the AI-IPR crisis is beyond question, but the past is full of examples of how the judiciary has functioned as a bridge in times of severe technological upheavals. In the absence of dedicated laws to guide the resolution of AI-generated IP conflicts, courts have a number of well-known doctrinal mechanisms at their disposal. The doctrine of work for hire, which automatically assigns copyright ownership to the employer for the creation of a work by an employee as part of such employment, can be extended through analogy to grant IP rights to the corporate employer that commissions, configures, and deploys an AI system that produces a particular kind of work. In some cases, the common law doctrine of unjust enrichment may offer a relevant basis for finding damages for rights holders who have their works used without permission to train AI, in absence of an express statutory cause of action.

In India, courts have proven to be very flexible in applying existing common law principles to new technology. The Delhi High Court's decisions on the liability of ISPs for copyright infringement and the Bombay High Cour's approach to digital rights in broadcasting law offer lessons to the courts on dealing with disruptive technology. Authoritative decisions on the questions of authorship, ownership and liability under AI could be obtained through a limited number of test cases sponsored by rights holders associations and carefully-crafted appellate proceedings, even without legislative interventions, giving a sense of a developing coherent domestic jurisprudence. In this way the judiciary is not just a silent partner in the statute but a dynamic participant, working within and interpreting the boundaries of the statute.

VI. Conclusion

The combination of Artificial Intelligence and Intellectual Property Rights is one of the biggest legal issues in the twenty first century. The current legal system, which is based on human creativity and individual authorship is structurally deficient to address an age of machines creating protectable IP products.

The paper has recommended the following measures: (i) establishment of a sui generis protection for AI-generated outputs; (ii) reform of the definition of authorship and inventorship in Indian and international laws; (iii) mandatory AI disclosure requirements for IP registrations; (iv) international harmonisation through a WIPO treaty; and (v) statutory fair dealing exception for AI training data. Combined, these measures would ensure legal certainty, promote innovation through AI, uphold human creativity, and enable a fair distribution of the benefits of AI.

It is not a matter of protecting AI or protecting humans — its a matter of a rational legal framework for AI creativity or for an ungoverned vacuum to remain. In Burrow-Giles Lithographic Co. v. Sarony, the Supreme Court of the United States stated that the law must change to cover ‘new forms' created through technology. That evolution has long been overdue and it is now.

Reference

1. World Intellectual Property Organization (WIPO), "Artificial Intelligence and Intellectual Property Policy," WIPO Publication No. 1055 (2019), https://www.wipo.int/publications/en/details.jsp?id=4455.

2. Ryan Abbott, "I Think, Therefore I Invent: Creative Computers and the Future of Patent Law," 57 B.C. L. Rev. 1079, 1082 (2016).

3. Copyright Act, 1957 (India), § 2(d)(vi); 17 U.S.C. § 101 (2018).

4. Andres Guadamuz, "Do Androids Dream of Electric Copyright? Comparative Analysis of Originality in Artificial Intelligence Generated Works," 2017 Intell. Prop. Q. 169, 171 (2017).

5. Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022).

6. Burrow-Giles Lithographic Co. v. Sarony, 111 U.S. 53, 58 (1884).

7. Feist Publications, Inc. v. Rural Telephone Service Co., 499 U.S. 340, 345 (1991).

8. Eastern Book Company v. D.B. Modak, (2008) 1 SCC 1 (India).

9. Naruto v. Slater, 888 F.3d 418, 420 (9th Cir. 2018).

10. U.S. Copyright Office, "Copyright and Artificial Intelligence, Part 1: Digital Replicas," 89 Fed. Reg. 38103 (May 6, 2024).

11. Trade Marks Act, 1999 (India), § 2(zb); 15 U.S.C. § 1051 (Lanham Act, United States); Trade Marks Act 1994 (UK), § 1.

12. See generally Stacey Dogan, “Brands and the Internet of Things,” 47 Seton Hall L. Rev. 1 (2016); Dinwoodie, Graeme B., “Trademark Law and the Internet,” in Research Handbook on the Law of Intellectual Property (Edward Elgar 2017).

13. Defend Trade Secrets Act, 18 U.S.C. § 1836 (United States); Information Technology Act, 2000 (India), § 43; cf. European Union Trade Secrets Directive (EU) 2016/943.

14. Mark Lemley & Bryan Casey, “Who Goes First? The Emerging Law of AI ‘Inventors,’” 2019 U. Chi. Legal F. 375 (2019); see also WIPO, “Artificial Intelligence and Intellectual Property Policy,” WIPO Publication No. 1055 (2019), at 23-26.

15. Stephen Thaler v. Comptroller-General of Patents, Designs and Trade Marks [2023] UKSC 49 (UK Supreme Court).

16. DABUS (Device for the Autonomous Bootstrapping of Unified Sentience), Patent Application No. PCT/US2019/053769 (2019).

17. Patents Act, 1970 (India), § 6; European Patent Convention, Art. 60(1).

18. Ryan Abott & Bret Bogenschneider, "Should Robots Pay Taxes? Tax Policy in the Age of Automation," 46 Harv. J. on Legis. 145, 160 (2018).

19. Computer Sciences Corp. v. United States, 50 Fed. Cl. 388, 399 (2001).

20. Google LLC v. Oracle America, Inc., 593 U.S. 1 (2021).

21. Authors Guild v. Google, Inc., 804 F.3d 202, 213 (2d Cir. 2015).

22. Andrei Iancu, "Artificial Intelligence: The Intellectual Property Response," Remarks at the Stanford Law School CodeX FutureLaw Conference (Apr. 2019).

23. Tencent Technology Co. Ltd. v. Shanghai Yingxun Technology Co. Ltd., Shenzhen Nanshan District People's Court, No. (2019) Yue 0305 Min Chu 14010 (2019).

24. Copyright Law of the People's Republic of China (as amended 2020), National Copyright Administration of China; CNIPA, “Guidelines for Patent Examination” (2023 Edition).

25. Yong Wan, “AI-Generated Works and Copyright Law: Exploring the Boundaries of Authorship in China,” 53 IIC 1097 (2022).

26. Thaler v Commissioner of Patents [2021] FCA 879 (Federal Court of Australia), per Beach J.

27. Thaler v Commissioner of Patents [2022] FCAFC 62 (Full Federal Court of Australia); see also IP Australia, “Patents, Trade Marks, Designs and IP Government Review: AI Issues Paper” (2023).

28. European Parliament Resolution of 20 October 2020 on Intellectual Property Rights for the Development of Artificial Intelligence Technologies, P9_TA(2020)0277.

29. Department for Promotion of Industry and Internal Trade (DPIIT), Government of India, "National IPR Policy (2016), https://dipp.gov.in/policies-rules-and-acts/policies/national-iprpolicy.

30. Yochai Benkler, "The Wealth of Networks: How Social Production Transforms Markets and Freedom" 2 (Yale Univ. Press 2006).

31. Jane C. Ginsburg & Luke Ali Budiardjo, "Authors and Machines," 34 Berkeley Tech. L.J. 343, 346 (2019).

32. Robert Denicola, ;Ex Machina: Copyright Protection for Computer-Generated Works69 Rutgers L. ev. 251, 272 (2016).

33. WIPO, "WIPO Conversation on IP and AI: Revised Issues Paper on IP and AI, Third Session, WIPO/IP/AI/3/GE/21/1 (2021).

34. Peter Lee, “Artificial Intelligence, Liability, and the Public-Private Divide,” 1 J.L. & Innovation 1 (2018); cf. Copyright Act, 1957 (India), § 51.

35. Andersen v. Stability AI Ltd., No. 3:23-cv-00201 (N.D. Cal. 2023); Getty Images (US), Inc. v. Stability AI, Ltd., No. 1:23-cv-00135 (D. Del. 2023).

36. Copyright Act, 1957 (India), § 51(a)(ii); cf. CBS Songs Ltd. v. Amstrad Consumer Electronics plc [1988] AC 1013 (HL) (UK authorisation doctrine).

37. Berne Convention for the Protection of Literary and Artistic Works, Art. 6bis (Paris Act 1971).

38. Copyright Act, 1957 (India), § 57; see Amar Nath Sehgal v. Union of India, 2005 (30) PTC 253 (Del HC).

39. Jane C. Ginsburg, “Human Authorship, Copyright, and the Digital Creative Environment,” in The Oxford Handbook of Intellectual Property Law (OUP 2018) 305, 318.

40. Carlos M. Correa, “Patent Evergreening in Developing Countries,” Research Papers No. 32 (South Centre 2011); Novartis AG v. Union of India, (2013) 6 SCC 1.

41. Ekins & Hopkins, “Using Deep Learning to Innovate on Lead Compounds,” 23 Drug Discovery Today 1287 (2018); see also Patents Act, 1970 (India), § 3(d).

42. Gillian Davies, Copyright and the Public Interest (2nd ed., Thomson Sweet & Maxwell 2002) 145-160; see Indian Performing Right Society Ltd. v. Eastern Indian Motion Pictures Association, (1977) 2 SCC 820.

43. UK Intellectual Property Office, “Artificial Intelligence and IP: Copyright and Patents: Government Response to Consultation” (February 2023).

44. See R.G. Anand v. Deluxe Films, (1978) 4 SCC 118 (India); Super Cassettes Industries Ltd. v. Myspace Inc., 236 (2017) DLT 478 (Del HC).

45. Shamnad Basheer & Annalisa Telesetsky, “Technology and IP Policy in Developing Countries: Navigating the Governance Gap,” 15 Chi.-Kent J. Intell. Prop. 367 (2016).