Navigating the Uncharted Waters of Copyright Protection for Machine-Generated Works
Executive Summary
The rapid proliferation of generative AI systems has created a profound legal vacuum in India's copyright framework. The January 2024 lawsuit by ANI (Asian News International) against OpenAI in Delhi High Court marks a watershed moment, forcing Indian courts to confront fundamental questions about AI-generated content, training data usage, and the boundaries of fair dealing. This article provides a comprehensive analysis of the legal landscape, examining Section 2(d)(vi) of the Copyright Act, 1957, the government's 2025 expert panel on Copyright Act reform, DPIIT's position on AI training, and the broader implications for content creators, technology companies, and legal practitioners.
Key Developments:
- ANI v. OpenAI: First major AI copyright case in India (Delhi HC, 2024)
- Section 2(d)(vi) ambiguity: No clear ownership framework for AI-generated works
- Government expert panel formed in 2025 to review Copyright Act
- DPIIT position: No blanket fair use exemption for commercial AI training
- Global developments: US Copyright Office, EU AI Act, UK approach
Table of Contents
- Introduction: The AI Copyright Crisis
- ANI v. OpenAI: The Landmark Indian Case
- Section 2(d)(vi) Copyright Act: The Authorship Question
- Government's 2025 Expert Panel on Copyright Reform
- DPIIT Position on AI Training and Fair Dealing
- Comparative Global Frameworks
- Delhi High Court Jurisprudence on Authorship and Originality
- AI and Related Rights: Performers, Moral Rights
- Practitioner's Analysis: Risk Matrix and Compliance
- Future Directions and Legislative Proposals
- Conclusion
1. Introduction: The AI Copyright Crisis
The emergence of large language models (LLMs) like GPT-4, Claude, and Gemini has fundamentally disrupted traditional copyright paradigms. These systems are trained on vast corpora of copyrighted material, generate outputs that may infringe existing copyrights, and produce works whose ownership status remains legally uncertain.
The Tripartite Problem
| Issue |
Description |
Legal Ambiguity |
| Input |
Training AI on copyrighted works |
Is it infringement or fair dealing? |
| Process |
AI models encode copyrighted patterns |
Are model weights derivative works? |
| Output |
AI-generated content resembling protected works |
Who owns AI outputs? Infringement? |
Statistical Overview
| Metric |
Value |
Source |
| Training data size (typical LLM) |
500+ billion tokens |
Industry reports |
| Percentage of copyrighted material |
60-80% estimated |
Research studies |
| Global AI copyright lawsuits (2023-25) |
40+ |
Legal databases |
| Indian AI copyright cases |
5+ |
Court records |
India's Copyright Framework - Key Provisions
| Section |
Content |
AI Relevance |
| Section 2(d) |
Definition of "author" |
No AI authorship provision |
| Section 2(o) |
Definition of "literary work" |
Computer programs included |
| Section 13 |
Works in which copyright subsists |
Requires human authorship? |
| Section 52 |
Fair dealing exceptions |
AI training not covered |
2. ANI v. OpenAI: The Landmark Indian Case
Case Overview
In January 2024, Asian News International (ANI), India's largest news agency, filed suit against OpenAI in the Delhi High Court, alleging massive copyright infringement through the training and outputs of ChatGPT.
Key Allegations
| Allegation |
Details |
| Unauthorized Training |
ChatGPT trained on ANI's copyrighted news content |
| Output Infringement |
ChatGPT reproduces ANI articles verbatim or substantially |
| Attribution Failure |
No credit to ANI for derived content |
| Commercial Use |
OpenAI profits from ANI's copyrighted work |
ANI's Claims
| Claim |
Legal Basis |
| Copyright infringement |
Section 51, Copyright Act, 1957 |
| Passing off |
Common law |
| Unfair competition |
Tortious liability |
| Violation of moral rights |
Section 57, Copyright Act |
OpenAI's Potential Defenses
| Defense |
Legal Basis |
Strength |
| Fair dealing for research |
Section 52(1)(a) |
Weak - commercial use |
| Transformative use |
Judicial interpretation |
Uncertain in India |
| Technical process argument |
Process vs. expression |
Untested |
| No substantial similarity |
Fact vs. expression |
Case-specific |
Significance for Indian Law
| Aspect |
Implication |
| Precedent-setting |
First major AI/LLM case in India |
| Training data question |
Will courts allow mass scraping? |
| Output liability |
When does AI output infringe? |
| Fair dealing scope |
Commercial AI training covered? |
3. Section 2(d)(vi) Copyright Act: The Authorship Question
Statutory Framework
Section 2(d) of the Copyright Act, 1957 defines "author" as:
| Clause |
Type of Work |
Author |
| (i) |
Literary/dramatic work |
The author of the work |
| (ii) |
Musical work |
The composer |
| (iii) |
Artistic work |
The artist |
| (iv) |
Photograph |
Photographer |
| (v) |
Cinematograph film |
Producer |
| (vi) |
Computer-generated work |
The person who causes the work to be created |
The AI Authorship Gap
Section 2(d)(vi) was inserted in 1994 to address computer-generated works, but it was designed for traditional software tools, not autonomous AI systems.
| Scenario |
Current Law |
AI Complication |
| Human uses Word to write |
Human is author |
Clear |
| Human uses AI with specific prompts |
Human causes creation? |
Ambiguous |
| AI generates with minimal prompts |
Who "caused" creation? |
Highly uncertain |
| Fully autonomous AI output |
No identifiable causer |
Legal vacuum |
Interpretive Challenges
| Question |
Possible Answers |
Implications |
| Who "causes" AI output? |
User, developer, AI company |
Ownership disputes |
| What level of human input required? |
Detailed prompts, fine-tuning, none? |
Threshold issues |
| Is AI a tool or creator? |
Tool analogy vs. independent creator |
Fundamental framework |
Judicial Interpretation Needed
Indian courts have not yet definitively ruled on Section 2(d)(vi) in the AI context. Key questions include:
- Causation Test: What constitutes "causing" the work to be created?
- Originality Requirement: Do AI outputs meet the "modicum of creativity" standard?
- Fixation: Is AI output "fixed" for copyright purposes?
- Expression vs. Idea: How to distinguish AI's learned patterns from expression?
In early 2025, the Ministry of Commerce and Industry announced the formation of an expert panel to review the Copyright Act, 1957, with specific focus on AI-related challenges.
Panel Composition
| Member Type |
Representation |
| Legal experts |
IP law specialists, retired judges |
| Industry representatives |
Tech companies, publishers, content creators |
| Academic experts |
IP scholars, AI researchers |
| Government officials |
DPIIT, MeitY officials |
Key Issues Under Review
| Issue |
Current Position |
Reform Proposals |
| AI Authorship |
Section 2(d)(vi) inadequate |
New definition needed |
| Training Data Use |
No clear exception |
TDM exception debate |
| Output Ownership |
Uncertain |
Registration system? |
| Liability Framework |
Traditional infringement |
AI-specific rules |
International Benchmarking
| Jurisdiction |
Approach |
Relevance for India |
| UK |
Computer-generated work = programmer as author |
Established precedent |
| USA |
Human authorship required (Copyright Office) |
Restrictive model |
| EU |
AI Act + Copyright Directive |
Regulatory approach |
| Japan |
Flexible TDM exception |
Pro-innovation model |
Expected Outcomes
| Deliverable |
Timeline |
Status |
| Consultation paper |
Q1 2025 |
Released |
| Stakeholder consultations |
Q2-Q3 2025 |
Ongoing |
| Draft recommendations |
Q4 2025 |
Expected |
| Legislative proposal |
2026 |
Anticipated |
5. DPIIT Position on AI Training and Fair Dealing
Official Stance
The Department for Promotion of Industry and Internal Trade (DPIIT) has taken a cautious position on AI training using copyrighted content.
Key Policy Positions
| Issue |
DPIIT Position |
| Blanket Fair Use for AI |
Not supported |
| Commercial AI Training |
Requires licensing |
| Research/Academic Use |
May qualify under Section 52(1)(a) |
| Text and Data Mining (TDM) |
No automatic exception |
Section 52 Fair Dealing Analysis for AI
| Exception |
Text |
AI Training Applicability |
| Section 52(1)(a) |
Fair dealing for private/personal use, criticism, review |
Not for commercial AI training |
| Section 52(1)(a)(i) |
Research |
Academic AI research may qualify |
| Section 52(1)(b) |
Reporting current events |
Limited applicability |
| Section 52(1)(h) |
Adaptation for computer program |
Narrow exception |
Licensing Framework Proposals
| Model |
Description |
Stakeholder View |
| Opt-out System |
Default permission with opt-out rights |
Tech industry preference |
| Opt-in System |
Explicit permission required |
Publisher/creator preference |
| Collective Licensing |
CMO-based mass licensing |
Compromise approach |
| Extended Collective Licensing |
Presumed representation |
Nordic model |
Industry Impact Assessment
| Sector |
Impact of Restrictive Approach |
Impact of Permissive Approach |
| News Media |
Revenue protection |
Innovation hindrance |
| Publishers |
Rights enforcement |
Market access concerns |
| Tech Companies |
Licensing costs |
Innovation enablement |
| Creators |
Attribution rights |
Exposure concerns |
6. Comparative Global Frameworks
United States
| Aspect |
Position |
| Copyright Office |
Human authorship required; AI outputs not copyrightable |
| Training Data |
Fair use defense available; litigation ongoing |
| Major Cases |
Getty v. Stability AI, NYT v. OpenAI, Authors Guild v. OpenAI |
European Union
| Regulation |
Content |
AI Relevance |
| AI Act (2024) |
Comprehensive AI regulation |
Disclosure requirements |
| Copyright Directive (2019) |
TDM exceptions |
Research vs. commercial distinction |
| Article 4 |
TDM for any purpose with opt-out |
AI training may qualify |
United Kingdom
| Feature |
Description |
| CDPA Section 9(3) |
Computer-generated works protected; author = programmer |
| Proposed TDM Exception |
Considered but withdrawn (2023) |
| Current Position |
Licensing-based approach preferred |
China
| Aspect |
Approach |
| Interim AI Measures (2023) |
Generative AI regulations |
| Training Data |
Must respect IP rights |
| Output Labeling |
Required for AI-generated content |
Comparative Summary
| Country |
AI Authorship |
Training Data |
Output Protection |
| India |
Uncertain (2(d)(vi)) |
No clear exception |
Uncertain |
| USA |
Not copyrightable |
Fair use litigation |
Not protected |
| UK |
Programmer as author |
Licensing preferred |
Protected |
| EU |
Human authorship required |
TDM exception |
Human element needed |
7. Delhi High Court Jurisprudence on Authorship and Originality
Foundational Principles
Indian courts have established key principles on authorship and originality that will inform AI copyright analysis.
A.R. Rahman v. Ustad Faiyaz Wasifuddin Dagar (2025)
Case Number: FAO(OS) (COMM) 86/2025
Court: High Court of Delhi
Judges: Justice C. Hari Shankar, Justice Om Prakash Shukla
Key Holding
The court held that performance does not equate to authorship under the Copyright Act:
"Section 2(d) defines the author of a musical work as the composer, and Section 17 vests copyright in the author. The court emphasized that originality requires independent creative contribution beyond traditional structures."
Relevance for AI Copyright
| Principle |
Application to AI |
| Authorship requires composition |
AI outputs require human creative contribution |
| Performance ≠ Creation |
Running AI ≠ authoring output |
| Originality needs independent contribution |
Mere prompting may be insufficient |
| Proof of creation required |
Cannot presume authorship from output |
Arun Chadha Copyright Case (2012)
Case Number: CS(OS) 1096/2009
Key Issue: Authors' special rights under Section 57
Holding
"The author's moral rights to claim authorship and restrain distortion survive assignment of copyright."
AI Implications
| Aspect |
Implication |
| Moral rights |
AI has no moral rights; only human authors |
| Attribution |
Human involvement required for attribution claims |
| Integrity rights |
Cannot distort AI output as no "author" to harm |
Originality Standards in India
| Case |
Standard Articulated |
| Eastern Book Company v. D.B. Modak |
"Sweat of the brow" insufficient; skill and judgment required |
| R.G. Anand v. Delux Films |
Ideas not protected; expression only |
| Star India v. Leo Burnett |
Substantial reproduction test |
Section 38 of the Copyright Act grants performers independent rights distinct from copyright in the underlying work.
| Right |
Section |
AI Application |
| Right to prevent fixation |
38(1) |
AI cannot be a "performer" |
| Right to prevent reproduction |
38(2) |
N/A |
| Moral rights of performer |
38A |
Human performers only |
AI Voice Cloning Issues
The proliferation of AI voice cloning raises significant concerns:
| Issue |
Current Law |
Gap |
| Voice replication |
Personality rights (tort) |
No statutory protection |
| Deepfakes |
IT Act Section 66D |
Limited to fraud |
| Attribution |
Section 57 moral rights |
Requires human author |
Moral Rights and AI
Section 57 - Author's Special Rights:
| Right |
Text |
AI Applicability |
| Paternity right |
Right to claim authorship |
Cannot apply to AI |
| Integrity right |
Right to prevent distortion |
No AI "honor" to protect |
The Human Authorship Requirement
Emerging consensus suggests:
| Position |
Rationale |
| Human authorship essential |
Copyright protects human creativity |
| Instrumental view of AI |
AI as tool, not creator |
| Incentive theory |
Copyright incentivizes humans, not machines |
9. Practitioner's Analysis: Risk Matrix and Compliance
Risk Assessment for Content Creators
| Risk |
Likelihood |
Impact |
Mitigation |
| AI training on content |
High |
High |
robots.txt, terms of use |
| Output plagiarism |
Medium |
High |
Monitoring, enforcement |
| Attribution loss |
High |
Medium |
Watermarking, registration |
| Revenue displacement |
High |
High |
Licensing, advocacy |
Risk Assessment for AI Companies
| Risk |
Likelihood |
Impact |
Mitigation |
| Training data litigation |
High |
Very High |
Licensing, documentation |
| Output infringement claims |
Medium |
High |
Filtering, disclaimers |
| Regulatory action |
Medium |
High |
Compliance programs |
| Reputational damage |
Medium |
Medium |
Transparency, attribution |
Compliance Framework for AI Training
| Step |
Action |
Legal Basis |
| 1 |
Document all training data sources |
Evidence preservation |
| 2 |
Obtain licenses where available |
Risk mitigation |
| 3 |
Implement opt-out mechanisms |
Best practice |
| 4 |
Maintain attribution systems |
Moral rights compliance |
| 5 |
Regular legal audits |
Ongoing compliance |
Compliance Framework for AI Users
| Step |
Action |
Purpose |
| 1 |
Verify AI output for originality |
Avoid infringement |
| 2 |
Add human creative elements |
Strengthen copyright claim |
| 3 |
Document creation process |
Prove authorship |
| 4 |
Check for third-party rights |
Due diligence |
| 5 |
Disclose AI involvement |
Transparency |
10. Future Directions and Legislative Proposals
Proposed Amendments to Copyright Act
| Proposal |
Description |
Support |
| New Section 2(d)(vii) |
Define AI-generated works |
Mixed |
| TDM Exception |
Add Section 52(1)(za) for research TDM |
Tech industry |
| Opt-out Rights |
Rights holder reservation mechanism |
Publishers |
| Registration System |
Voluntary AI output registration |
Moderate support |
Model Legislative Framework
Proposed Section 2(d)(vii):
"In relation to any work generated by an artificial intelligence system with substantial human creative input, the person who provides such creative direction and input shall be deemed the author."
Proposed Section 52(1)(za):
"The reproduction of a work for the purposes of text and data mining for scientific research by research organizations and cultural heritage institutions, provided that the work is lawfully accessible and the reproduction is not used for commercial purposes."
Industry Self-Regulation
| Initiative |
Description |
| Content Authenticity Initiative |
Digital watermarking for provenance |
| Partnership on AI |
Ethical AI development |
| C2PA Standard |
Content credentials for AI outputs |
Expected Timeline
| Year |
Expected Development |
| 2025 |
Expert panel recommendations |
| 2026 |
Draft Copyright Amendment Bill |
| 2027 |
Parliamentary consideration |
| 2028 |
Potential enactment |
11. Conclusion
The intersection of AI and copyright law represents one of the most significant legal challenges of our time. The ANI v. OpenAI case will be a bellwether for how Indian courts approach these issues, while the government's expert panel deliberations will shape legislative reform.
Key Takeaways
| Issue |
Current Status |
Likely Direction |
| AI Authorship |
Legally uncertain |
Human-centric approach likely |
| Training Data |
Infringement risk |
Licensing frameworks emerging |
| Fair Dealing |
Not clearly applicable |
Narrow exceptions possible |
| Output Protection |
Uncertain |
Human input requirement likely |
Recommendations for Stakeholders
| Stakeholder |
Recommendation |
| Content Creators |
Implement technical protections; engage in policy advocacy |
| AI Companies |
Develop licensing relationships; document compliance |
| Legal Practitioners |
Monitor developments; advise on risk management |
| Policymakers |
Balance innovation with creator rights |
The Path Forward
India stands at a crossroads. The choices made in the coming years will determine whether the country becomes a leader in responsible AI development or faces a fragmented, litigation-driven landscape. A balanced approach that:
- Protects human creativity and livelihoods
- Enables beneficial AI innovation
- Provides legal certainty for all stakeholders
- Respects international obligations
...will serve India's interests best as it navigates the AI revolution.
Key Citations
| Case/Document |
Citation |
Key Principle |
| ANI v. OpenAI |
Delhi HC (2024) |
AI training copyright issues |
| A.R. Rahman v. Dagar |
FAO(OS)(COMM) 86/2025 |
Authorship vs. performance |
| Eastern Book v. D.B. Modak |
(2008) 1 SCC 1 |
Originality standard |
| Copyright Act, 1957 |
Section 2(d)(vi) |
Computer-generated works |
| DPIIT Position Paper |
2025 |
No blanket fair use for AI |
Sources
- ANI v. OpenAI, Delhi High Court (2024) - Court filing
- A.R. Rahman v. Ustad Faiyaz Wasifuddin Dagar, FAO(OS)(COMM) 86/2025, Delhi High Court (24-09-2025)
- The Foundry Visionmongers v. Singarajan VFX, CS(COMM) 461/2022, Delhi High Court (05-05-2025)
- Copyright Act, 1957 (as amended)
- DPIIT discussion paper on AI and Copyright (2025)
- Lexology: "Government formed expert panel in 2025 to review Copyright Act"
- Iplink-asia: "Section 2(d)(vi) analysis"
- US Copyright Office: AI and Copyright Guidance (2023-24)
- EU AI Act (2024)
