Introduction
The Parliament passed the Copyright Act, 1957 on June 4, 1957, (Act No. 14 of 1957). Presidential assent came the same day. The Act itself did not come into force until January 21, 1958, after the Central Government issued the requisite notification in the Official Gazette.
So the Copyright Office's annual commemoration on June 4 marks enactment, not commencement. A technicality, yes. But the Copyright Office knows this distinction. They chose the enactment date anyway, because June 4 is when Parliament acted. The seven-and-a-half months of delay before commencement were administrative.
Having trouble understanding Copyright law? Click here
This year, June 4 fell on a Wednesday. The Copyright Office held a virtual session and the theme was "Foundation, Evolution and Future Horizons of Copyright Law."
You can read that theme as ceremonial or take it as an acknowledgment, from the institution that administers the statute, that the "horizons" portion is crowded in a way it has not been since 1994. The Delhi High Court reserved judgment in ANI Media Pvt. Ltd. v. OpenAI Inc. & Anr. (CS(COMM) 1028/2024) on March 27, 2026, after hearing the matter in full. Stephen Thaler's copyright application for an AI-generated artwork, filed in India in 2022, just got a Delhi High Court direction to be decided within eight weeks. The DPIIT's expert committee published its Working Paper on Generative AI and Copyright in December 2025. Section 52 is the fair dealing provision that nobody outside IP circles had heard of five years ago, and it is now the clause that will determine whether India's AI sector can train on copyrighted data without a licensing agreement.
That is a lot happening around a statute that received its last amendment in 2012.
What Section 52 Actually Says, and Why It Matters Right Now
Before going back to 1957, go to Section 52. It is the provision that decides the ANI v. OpenAI case, or at least the one OpenAI is betting on.
Section 52 lists the acts that do not constitute infringement. The list is exhaustive. The Delhi High Court confirmed this in Super Cassettes Industries Ltd. v. Chintamani Rao, Section 52 is not illustrative. Courts cannot add to it, however pressing the technological need. The listed purposes are private use, research, criticism or review, news reporting, and certain educational and library uses. That is the complete set.
Now, when an AI company downloads a news agency's entire archive to train a large language model, which of those categories covers it? Not private use. Commercial deployment is the objective. Not criticism or review. Not news reporting. Not educational use in any statutory sense.
The US Copyright Act's Section 107 works differently. It sets out four factors (purpose of the use, nature of the work, amount taken, market effect), and courts apply them case by case. It is an open-ended standard, not a closed list. Google LLC v. Oracle America Inc., decided by the US Supreme Court in 2021, applied exactly this analysis and found that Google's copying of Java API code for Android was fair use. You could not get to that result under Section 52 of the Indian Copyright Act.
OpenAI's legal team knows this. In the ANI proceedings, one of the two amici curiae argued that storing copyrighted data for AI training might fall within Section 52(1)(a), the storage exception, provided no expressive content reaches the public. The other amicus took the opposite position: unlicensed commercial copying for AI training falls outside fair dealing and constitutes infringement regardless of what gets published downstream. Those two positions represent the outer limits of the argument. The Delhi High Court has to choose between them, or carve something in the middle.
The DPIIT Working Paper, published December 8, 2025, was clear that amending or expanding Section 52 will not solve the problem. Too narrow a revision risks being obsolete before it clears Parliament. Too broad a revision risks gutting the protection the provision exists to provide. The committee's preferred alternative is a hybrid compulsory licensing model where AI companies pay into a centralised fund for the right to train on lawfully accessed content. It is a workaround.
What the First Independent Copyright Law Was Trying to Do
Before 1957, India had the Indian Copyright Act, 1914. That Act was the Imperial Copyright Act, 1911, a UK statute applied to India with minor modifications. The modifications were not the fit to the Indian context, rather, it was British law dressed in Indian administrative clothing.
The 1957 Act was the first copyright legislation India actually wrote for itself. It replaced the 1914 Act, aligned India with the Berne Convention and the Universal Copyright Convention, established the Copyright Office and the Copyright Board, and set a protection term of life of the author plus 50 years. It ran to 79 sections across chapters.
One thing the drafters did that matters for the current AI debate: they wrote Section 2(d) broadly. "Author" means the person who creates the work. For computer-generated works, a category the 1957 drafters did not contemplate but which the 1994 amendment introduced, Section 2(d)(vi) says the author is "the person who causes the work to be created."
That phrase will determine the outcome of the Thaler application currently before the Registrar of Copyrights. Thaler caused his AI system DABUS to create A Recent Entrance to Paradise. Does that make Thaler the author under Section 2(d)(vi)? Or does the phrase require a person who directed a specific creative output, not merely a person who built and activated the machine that produced it?
The Copyright Office asked Thaler in July 2023 to identify a "natural person" as the author. Thaler refused. The application sat pending for nearly four years until Justice Tushar Rao Gedela directed resolution within eight weeks in April 2026. The Registrar's forthcoming decision will be the Copyright Office's first public position on AI authorship under Indian law. It will be reviewable, contestable, and almost certainly appealed whichever way it goes.
Also read: Can Restaurants Play Copyrighted Music?
Six Amendments in 69 Years: The Real Story Behind the Chronology
The six amendments (1983, 1984, 1992, 1994, 1999, 2012) get listed as progressive toward a modern statute.
- 1983: Paris Revision Act 1971 of the Berne Convention, Compulsory licensing for broadcasts and a Procedural cleanup.
- 1984: VHS cassettes arrived. Video piracy was undercutting the film industry in ways the 1957 Act could not touch. The 1984 amendment added anti-piracy provisions for video films and sound recordings.
- 1992: Life plus 50 years of protection now became life plus 60 years.
- 1994: The most forward-looking of the pre-2012 amendments. Computer programmes entered the definition of literary works for the first time. Performers' rights for singers, musicians, and actors appeared in the statute for the first time, allowing them to control the recording and broadcast of their performances. Satellite broadcasting had arrived in India in 1991 and 1992, and the broadcasting provisions needed updating to match. Section 2(d)(vi), the computer-generated works provision was also added.
- 1999: TRIPS alignment following WTO accession.
- 2012: The biggest one of all. Technological protection measures, anti-circumvention rules, statutory licensing for cover versions, broadcasting and the most consequential change for the music industry was the provision that authors cannot assign away their right to receive royalties from certain uses, even if they have otherwise assigned copyright. The 2012 amendment changed this arrangement giving them royalties despite assignment.
After 2012? Nothing. Huge improvements in technology did arrive like the OTT revolution, the streaming wars, the modern day social-media, user-generated content at scale, deepfakes, generative AI but no amendment to the Copyright Act.
Now, lets take a step back and compare that to what was happening internationally. The EU passed the Digital Single Market Directive in 2019, which includes a text and data mining exception that permits research organisations to use copyrighted works for computational analysis. Japan, introduced a broad data mining exception in 2018. Singapore amended its copyright statute comprehensively in 2021 with a specific computational data analysis exception. The UK has been consulting on a similar provision since 2022.
ANI v. OpenAI: Four Questions the Court Has to Answer
ANI filed the Commercial Suit in November 2024. The allegation was that OpenAI used its copyrighted news content to train ChatGPT, without license. ChatGPT also produced outputs that ANI attributed to itself, i.e., fabricated stories carrying the ANI byline. The relief sought was Rs. 2 crore in damages and an injunction. Four questions sit before Justice Amit Bansal, who heard the case before reserving judgment on March 27, 2026.
- Does storing copyrighted news data for AI training infringe the Copyright Act?
- Does using that data to generate user responses infringe?
- Does any of this fall within Section 52's fair dealing exceptions?
- Do Indian courts have territorial jurisdiction over OpenAI, a foreign defendant, at all?
If the Court declines jurisdiction, questions one through three stay unanswered. The Section 52 debate returns entirely to the legislature. The DPIIT Working Paper becomes the only game in the space and Parliament will have to act without a judicial anchor.
The judgment, when it arrives, will be the first time an Indian court formally states whether AI training on copyrighted content is lawful. That is not a small thing. But it will not be the last word.
The DC Circuit in Thaler v. Perlmutter confirmed in March 2025 that human authorship is a bedrock requirement under US copyright law. The US Supreme Court declined to review that ruling in March 2026. China's Beijing Internet Court granted copyright to an AI-generated image in 2023. No two major jurisdictions have landed in the same place. India, when the ANI judgment comes, will add its own answer to a question the world is still arguing about.
What the 2012 Amendment Left Unfinished
The 2012 amendment brought India into the digital era, at least as that era looked in 2012. Technological protection measures, anti-circumvention provisions, statutory licensing, strengthened royalty rights. All real achievements but it did not integrate the Copyright Act with the intermediary liability regime.
India's intermediary liability framework exists under Section 79 of the Information Technology Act, 2000, and the Intermediary Guidelines made under it. The Copyright Act has no equivalent of the US Digital Millennium Copyright Act's Section 512, the takedown-and-notice regime that gives rights holders a procedural mechanism to get infringing content removed from platforms, and gives platforms a procedural safe harbour if they comply.
Recently, when Hindware Ltd. sought to hold Google liable for auctioning its trademark as an AdWords keyword (a case already covered in this blog), the Court had to work through Section 79 of the IT Act, not the Copyright Act, to reach Google's platform liability. The two regimes interact but are not designed to work together.
Fourteen years have passed since the last amendment. The intermediary liability gap the 2012 Act left open is now the gap through which every AI copyright dispute partially falls.
The Thaler Question: Simpler Than It Looks
Let's step back from ANI for a moment, because the Thaler application raises a different question, and a cleaner one. Thaler did not prompt DABUS. He built the system, activated it, and the system produced A Recent Entrance to Paradise autonomously. Thaler then filed for copyright registration listing the AI as the sole author.
Every jurisdiction that has examined this, has denied protection. The US Copyright Office rejected the application. The DC Circuit affirmed. The Supreme Court declined to intervene. The UK Intellectual Property Office rejected the parallel application. Australia's Full Federal Court too, rejected it.
India's Copyright Office issued a discrepancy letter in July 2023 asking for a natural person to be named as author. Thaler refused. The application sat unresolved until April 2026, when Justice Tushar Rao Gedela of the Delhi High Court directed a decision within eight weeks. The Court said nothing about the merits of it.
Here is the tension in Indian law that makes this case more interesting than it might first appear. Section 2(d)(vi) says the author of a computer-generated work is "the person who causes the work to be created." This language was influenced by the UK's Copyright, Designs and Patents Act 1988, Section 9(3). It was for outputs of deterministic software, a programme that generates a specific table or a specific map when given a specific input. Thaler's DABUS is not that. It is a system that generates novel outputs without specific prompting.
Does "causes the work to be created" cover the person who built and ran the system, even when the specific creative output was not directed? The Indian Copyright Office will now decide this. The answer matters far beyond a single application as it sets the framework for every claim of authorship over AI-generated content.
Conclusion
The Copyright Office chose "Foundation, Evolution and Future Horizons" as its theme for a reason. At 69, the Copyright Act, 1957 has a foundation that mostly holds, an evolutionary record that shows real responsiveness to pressure when pressure became unavoidable, and horizons that are genuinely unsettled. The ANI v. OpenAI judgment, when it arrives, will answer some questions and leave others open. The Thaler decision from the Registrar will stake out the Copyright Office's position on AI authorship, but it will almost certainly be challenged. Section 52 will either be amended or it will not, and if it is not, courts will have to find workarounds that the text does not cleanly support.
What I find difficult to defend is the fourteen-year gap between 2012 and now. Not the absence of an AI-specific amendment before AI became a live issue. That would definitely be unreasonable to expect. But the failure to address the intermediary liability integration after 2012, or to build a notice-and-takedown regime into the Copyright Act, or to commission serious legislative work on digital copyright before a lawsuit forced the question. That reflects a legislative posture that responds to crises rather than anticipating them.
The creative industries (film, music, publishing, software) are substantial economic contributors. The AI sector is a stated policy priority. Designing a copyright framework that serves both interests is not an impossible task. Other jurisdictions have made genuine progress. Japan's 2018 data mining exception and the EU's 2019 DSM Directive are not perfect, but they gave their respective AI sectors a degree of legal clarity before the litigation arrived. India is still waiting for the litigation to provide that clarity.
The June 4 session marked 69 years but did not mark a solution. Only time will determine the future of AI & Copyright law.
Also read: Whether work needs to be original for a copyright?
~ Adv. Koushik Chittella
© All Rights Reserved
Disclaimer: None of the contents of this post constitute legal advice.