Artificial Intelligence and Copyright: A Comparative Analysis of India, China, and the United States

Introduction

Over the past decade, Artificial Intelligence (AI) has redefined human technology interaction and driven significant advancements across sectors such as healthcare, transportation, education, finance and more. With its rapid growth across various industries, AI has addressed complex real-world challenges through data-driven insights and automation. Further expanding its capabilities and gaining worldwide recognition. However, with its rapid expansion, AI has simultaneously raised serious Intellectual Property Concerns, posing a threat to human creators as they are accused of violating the IPR of human creators. These AI models are trained on massive datasets that may include copyrighted works, which raises concerns about unauthorised use and infringement. This has intensified a global debate over consent, attribution, liability of AI-generated works, who owns the AI-generated output work and whether the use of copyrighted works as training data for an AI system constitutes infringement or is considered transformative.

This emerging tension can be seen in the recent dispute between The Walt Disney Company and Chinese technology company ByteDance. The Walt Disney Company issued a cease-and-desist notice to ByteDance over its Seedance 2.0 application, reportedly enabling “virtual smash and grab” recreations of copyrighted characters from the Marvel and Star Wars franchise. Such controversies highlight the structural gaps of the existing framework to address AI-led creation and reproduction.

This blog will examine the contemporary legal framework around AI training on copyrighted material in India. Further, the blog will analyse the approaches of China and the United States to evaluate whether existing doctrines can create a balance between Innovation and the protection of human creativity.

AI Training and Copyrighted Inputs

The AI models are trained on huge datasets, which are scraped from the internet, by learning statistical patterns in human-produced text and images to approximate underlying data distributions. Over successive training cycles, these models are fine-tuned into sampled outputs, where the new generations estimate and replicate these patterns from prior data. By underlying these data distributions, AI learns the statistical patterns in human-created text and images. It also raises copyright concerns when protected works are “innovated” without consent and “reproduced” in expressive forms. Although developers often rely on defences like “fair use”, India, with its copyright law narrower than the fair dealing framework, demands that such use must not replicate the original expression or create a substitutive effect, thereby placing AI training practices under increasing legal scrutiny.

Copyright Protection under Indian Law

In India, the Copyright Act, 1957, has established safeguards against infringement and unauthorised use of protected works. Section 13 of the Act grants copyright protection to original literary, musical, artistic, and other works by human creativity. The anthropocentric organisation is also a dogma of the AI - created works because the majority of them are prepared using the existing data, not being written by humans. In order to qualify under this provision, the work must meet the “modicum of creativity” standard set in Eastern Book Co. v. D.B. Modak (2008). In this case, it was ruled that a minimal but substantive degree of human creativity is required. The challenge lies in determining whether the AI-generated content satisfies the requirement, as the human involvement here is restricted to the input of the prompt. Further, Section 2(d)(vi) attributes authorship in computer-generated works to the person who “causes the work to be created.” It does not clearly lay down the required threshold of creative contribution under copyright law if the case involves minimal prompting.

In addition, AI is not a legal entity according to Indian laws. Consequently, the current copyright system finds it difficult to cope with emerging concerns and issues, where the creator of content is not a human being, nor a legally recognised person. Since Section 51 of the Act considers infringement as an act committed by a “person” attributing the liability or ownership to an AI system, it remains a legal challenge to address. This ambiguity clearly highlights that the framework is not adequate and is not designed to address the current concerns.

Fair Dealing and AI Training in India

A main question that arises is whether AI training practices fall within the meaning of “fair use” under Section 52 of the Copyright Act, 1957. In India, the doctrine of fair use is limited to only such uses as creating private use, research, criticism, and reporting. As a result, the dispute between ANI Media Pvt. Ltd. and OpenAI raises the question of whether the large-scale use of copyright material for AI training can be qualified as “research” under fair use. ANI’s allegation of unauthorised use of its copyright content by OpenAI weakens the defence of fair use, especially when AI training involves systematic reproduction of authorised work.

Therefore, this case represents a significant judicial test of whether AI training practices can be included within India’s statutory framework. Different copyright scholars discuss that the fair use defence is weak in the circumstances of AI-written material. Indian jurisprudence emphasises that fair dealing is not reproduction or even similarity of original material. By this, the transformative defence is debatable as unsustainable when the output produced by applying AI copies or has a close relation to a work that is being protected. Thus, a legal issue is whether the use of AI training founded on the usage of copyrighted content may be regarded as fair use. It will be challenging to keep AI training sustainable with the help of the Act, based on the limited statutory environment in India.

Comparative Approaches : China and the United States

In a landmark ruling, the Hangzhou Intermediate Court in China examined whether the use of copyrighted work to train a generative AI system could qualify as fair use. Although Article 24 of the 2020 Chinese Copyright Law doesn’t expressly mention AI training, its flexible fair use provision allows for an interpretation. The court stated that AI training on copyrighted inputs may be acceptable if there is no intention to reproduce, interfere, or cause unreasonable harm to the original work or to the legitimate interests of the copyright holder.

This is a framework that the court has chosen to take a comparatively lenient approach toward the use of input data and the AI training because of the large datasets that are used in machine learning, and has placed stricter scrutiny upon AI-generated content that has the potential to violate or displace original works. In addition, China has also strived to overcome the obstacles of the spread of AI in the context of its remaining copyright law. The Copyright Law of the People’s Republic of China does not specifically comment on the concept of AI-generated works; it still gives a wide framework, which can potentially protect the owners of the copyrighted materials. This shows China’s broader policy objective, where innovation is encouraged without undermining the core principle of copyright protection.

The copyright law followed by China puts an emphasis on the human authorship and originality since the piece of work involves a direct connection to human creativity. In the event that there is any suggestion that AI tools have been applied to the creative process, the courts evaluate whether or not there is a substantive human input in place, which may be the selection, arrangement or editing in order to be able to influence the final product. Without such a human intervention, the output might not be perceived as a working need to be protected. Chinese courts have clarified through cases such as, such as, Feilin v. Baidu and Shenzhen Tencent v. Shanghai Yingxun, which demonstrate that copyright protection is granted only when there is a meaningful human contribution. In the past, copyright has been granted by courts because the users demonstrated a high level of creativity by being quick in design and editing after the creation. The Zhangjiagang court went further to refuse to offer protection when it was established that there was little human input. Together, from these rulings, it is clear that China is moving forward with a balanced approach, where copyright is available for AI-assisted work, but only when human creativity is identified.

In the United States, the copyright debate over AI training has reached a critical stage, as courts are divided on whether copying copyrighted works to train a generative AI system qualifies as fair use, as codified in Section 107 of the Copyright Act of 1976. Technology companies argue that AI training is “transformative” because it converts copyrighted material into new outputs. However, recent rulings show judicial disagreement. The U.S. Copyright Office has stated that training AI models involves prima facie copying of protected works, which is why legal justification for fair use is needed. Even if some AI training uses may be considered as transformative, the determination will heavily depend on factors such as the purpose of the use, the source of the material, and the effect it has on the market.

Under U.S. copyright law, the protection extends to original works of authorship that are fixed in a tangible medium, with ownership typically vesting in the human author. The controversy surrounding Zarya of the Dawn highlighted the copyright office’s position that purely AI-generated works lack protectable authorship. They even proposed rules clarifying that works containing AI-generated material may be protected if there is sufficient human creative output. At the same time, major lawsuits like Andersen v. Stability AI and The New York Times v. OpenAI has also raised critical questions about whether the use of copyright material for AI training constitutes fair use or infringement. With growing AI-related copyright disputes, U.S. legislation has proposed the Generative AI Copyright Disclosure Act of 2024 and the No AI FRAUD Act, which seek to enhance transparency in AI training datasets and prevent unauthorized AI based impersonation.

In Authors Guild v. Google Inc., the U.S. Second Circuit held that digitising books to create a searchable database was transformative because it served a new function and didn’t substitute the original work. However, there is still uncertainty on whether this reasoning applies to generative AI systems, which are also trained on a vast input database and produce outputs resembling the original material. Additionally, the U.S. copyright office maintains that copyright protection is given only to works of human creativity, excluding the content generated by AI, thereby reinforcing the importance of human authorship in copyright law.

The way forward

The growing conflict between AI innovations and copyright protection requires the introduction of legal statutes on AI training and authorship, as well as liability. Laws need to specify how far copyrighted material may be utilised in the training of AI and set the bar for significant human input in the outputs produced by AI. It is necessary to introduce transparency in training AI and controlled licensing practices, as this will contribute to finding the balance between rights owners and the technological process. AI has become a global phenomenon, and therefore, global collaboration will be necessary to have a uniform and harmonised global copyright system.

Conclusion

The intersection of artificial intelligence and copyright law is one of the most debated issues of jurisdictional difficulty in the digital age. With the examination by the Indian framework that is covered by the Copyright Act, 1957, the current provisions are not only anthropocentric, but they also find it challenging to integrate AI-driven creation and training processes. Comparative practice of China and the United States indicates the changing and ambiguous meanings, especially in the domain of fair use, authorship, and transformative use. The courts struggle to reconcile innovation and protection of human creativity; there are legal grey areas, notably in the case of large-scale AI training based on a copyrighted feed.

Author: Samikshya Rout, in case of any queries please contact/write back to us via email to chhavi@khuranaandkhurana.com or at  Khurana & Khurana, Advocates and IP Attorney.

References

1. The Evolution and Future of Artificial Intelligence | CMU’ https://www.calmu.edu/news/future-of-artificial-intelligence

2. Shireen Yachu, ‘A Look at Generative AI in Terms of Intellectual Property Rights’ (Vidhi Centre for Legal Policy, 19 March 2025) https://vidhilegalpolicy.in/blog/a-look-at-generative-ai-in-terms-of-intellectual-property-rights/

3. ‘Artificial Intelligence and Intellectual Property’ https://www.wipo.int/en/web/frontier-technologies/artificial-intelligence/index

4. ‘Seedance: ByteDance to Curb AI App after Disney Legal Threat’ (16 February 2026) https://www.bbc.com/news/articles/c93wq6xqgy1o

5. ‘Section 13 in The Copyright Act, 1957’ https://indiankanoon.org/doc/4010217/

6. Eastern Book Co. v. D.B. Modak (2008) 1 SCC

7. The Copyright Act, 1957, Section 2(d)(vi)

8. The Copyright Act, 1957, Section 51

9. ANI Media Pvt. Ltd. and OpenAI 2024

10. Centre for Advanced Studies in Cyber Law and AI CASCA, ‘ANI v. OpenAI: The Intersection of Copyright and Artificial Intelligence in India’ (CASCA, 4 February 2025) https://www.cascargnul.com/post/openai-v-ani-the-intersection-of-copyright-and-artificial-intelligence-in-india

11. Copyright Law of the People's Republic of China, Section 24, ‘WIPO Lex’ https://www.wipo.int/wipolex/en/text/466268

12. Xinhang He and Pingji Shan, ‘China’s Regulations on the Attribution of AI-Generated Content: An Exploration Based on the Open-Ended Approach’ (2025) 20 Journal of Intellectual Property Law & Practice 318 https://doi.org/10.1093/jiplp/jpae109

13. Ju Yoen Lee, ‘Artificial Intelligence Cases in China: Feilin v. Baidu and Tencent Shenzhen v. Shanghai Yingxin’ (2021) 7 China and WTO Review https://cwto.net/index.php/CWR/article/view/180

14. Barbara Li, ‘Global AI Governance Law and Policy: China | IAPP’ (IAPP.org, 12 November 2025) https://iapp.org/resources/article/global-ai-governance-china

15. Copyright Law United States 1976, Section107

16. Blake Brittain and Blake Brittain, ‘AI Copyright Battles Enter Pivotal Year as US Courts Weigh Fair Use’ Reuters (5 January 2026) https://www.reuters.com/legal/government/ai-copyright-battles-enter-pivotal-year-us-courts-weigh-fair-use-2026-01-05/

17. ‘Andersen v. Stability AI: The Landmark Case Unpacking the Copyright Risks of AI Image Generators – NYU Journal of Intellectual Property & Entertainment Law’ https://jipel.law.nyu.edu/andersen-v-stability-ai-the-landmark-case-unpacking-the-copyright-risks-of-ai-image-generators/

18. ‘“Fair Use” in the Age of AI | India Corporate Law’ https://corporate.cyrilamarchandblogs.com/2025/04/fair-use-in-the-age-of-ai/