ADR in Resolving Copyright AI Training Data Disputes: Analysing the DPIIT Working Paper's 'One Nation, One License' Philosophy

Abstract

The DPIIT Working Paper on Licensing Generative AI (December, 2025), called One Nation, One License, One Payment, sets out to create a mandatory universal licensing framework- managed by a designated collecting agency (CRCAT), to regulate the use of copyright protected materials as input data for generative AI systems. Despite its ambitious

design, the framework suffers significantly from its lack of provisions regarding dispute resolution. This article will explore that shortcoming and will propose that Alternative Dispute Resolution (ADR) methods: especially structured mediation as per the Mediation Act, 2023 and arbitration pursuant to the Arbitration and Conciliation Act, 1996, along with hybrid mediation-arbitration regimes, provide a superior means of resolving all of the potential disputes that could arise within any universal licensing structure. Also provided in the paper are statistical data (from the WIPO Arbitration and Mediation Centre showing a

70% increase in IP-related disputes in 2025 with copyright and digital content making up 71% of those disputes), comparative data from the EU, UK, US, Japan and WIPO and an overview of Entrant’s current IP-ADR environment as established through the Boudhik Smadhan Program. To address this gap, the article will also propose a specialized ADR panel for IP disputes within the Copyright Office; model ADR clauses, fast-track arbitration rules and methods for determining royalties will be provided. These proposed recommendations are meant to be enforceable within India and its legal architecture while

also staying in long-run view as well to adapt to the ever changing landscape of AI governance.

Keywords: Generative AI Licensing, Alternative Dispute Resolution (ADR), DPIIT Working Paper, Copyright-AI Disputes, Mediation Arbitration Hybrid

I. Introduction

A significant area of debate within the realm of global intellectual property governance has arisen at the nexus between generative (creation-based) artificial intelligence (AI) technologies and copyright law. This is particularly evident in the case of India, where the

creative economy provides employment for millions and the AI industry is slated to play a major role in helping the country achieve its economic growth goals. With the release of its 125-page Working Paper by the Department for Promotion of Industry and Internal Trade (DPIIT) in December 2025, the Indian government's commitment to regulation of AI technologies has become abundantly clear, India will not wait for the judicial system to craft the necessary legal precedents; instead, they will create a comprehensive legislative framework governing access to AI training data.

The central proposal found in the Working Paper is a mandatory, universal licensing structure governed by collecting agencies (or Collecting Royalties for Copyrighted-Accessed Works for AI Technologies (CRCAT)) that will allow all AI developers to pay a pooled

(royalty-based) license fee for the right to use any copyright protected work that has been lawfully accessed for training purposes. While innovative, the absolute nature of this proposal is found to be highly contentious. Many legitimate concerns have been raised, ranging from

the lack of text-and-data mining (TDM) provisions in the Copyright Act of 1957 (in other words, no TDM exception exists), to the unavailability of individual licenses at the level that will be needed in order to conduct AI training, and how this system will be administered in practice. 2 There are numerous independent authors, publishers, and AI developers who have expressed differing but equally valid concerns with respect to the proposal's structure, its Constitutionality, and how it will be successfully administered.

Dispute resolution is a significant area that was not adequately addressed in the Working Paper, despite recognizing that there will likely be disputes resulting from licensing terms

related to royalty rates, distribution to rights-holders, and enforcing licensing terms. The paper provided only vague references to "institutional mechanisms; and "judicial review,"

and did not adequately consider that the anticipated volume and variety of potential disputes, or the complexity of the technical disputes resulting from the use of AI, are not well suited to conventional litigation.

This paper contends that Alternative Dispute Resolution (ADR) (i.e., mediation, arbitration, expert determination and hybrid med-arb) provides the most appropriate means of resolving the disputes anticipated to arise under any mandatory AI licensing regime. Section II provides a critical review and analysis of the DPIIT Working Paper's proposals and identifies the dispute resolution gap. Section III reviews and evaluates the suitability of specific ADR mechanisms for copyright-AI disputes. Section IV provides a practical ADR framework for

India. Section V analyzes relevant international examples and makes recommendations. Section VI provides a series of policy recommendations.

II. Overview and Critical Analysis of the DPIIT Working Paper

2.1 Core Proposals

There is a basic conflict between the fact that to train an AI system, you need to ingest huge volumes of material protected by copyright and that since it will not be possible to get the consent of every single author of that material prior to use, that means that the AI developer

will be committing copyright infringement as defined in Section 14 of the Copyright Act, 1957. The Working Paper suggests that this conflict can be resolved through a statutory structure called "One Nation, One License, One Payment an example of the proposed

statutory structure is as follows:

1. AI developers will be able to access and use all licensed copyrighted materials for AI training without the permission of individual copyright owners;

2. All revenues generated from the use of licensed copyrighted materials by AI developers will be collected by a single designated not-for-profit organization that will serve as the copyright collectives and pay royalties to copyright owners;

3. Unlike the EU AI Act, which allows some copyright holders to refuse use of their works, there will not be a similar opportunity for copyright holders under the copyright remuneration collective to opt-out of paying royalties;

4. The rates of royalties paid to copyright owners will be based on a delineated institutional methodology to be established by the Working Paper, which does not contain any detailed descriptions of what that methodology should be;

5. Transparency and accountability should be included as part of the governance structure of the CRCAT.

2.2 Strengths

Crediting the proposal for identifying the absence of a TDM exception in the Copyright Law as a disregard for traditional legal reasoning and rejecting the possibility of fair dealing in this context may, ultimately, distance the parties. With a single-window structure, good governance will help reduce transaction costs for AI developers and rights holders. The obligation of remuneration will mitigate the potential risk of circumvention of Indian Copyright Law via extraterritorial training, since many countries with similar exceptions also allow TDM use.

2.3 Critical Weaknesses

Critics of the Working Paper (including some who are authors of articles for SpicyIP, and have submitted formal comments to DPIIT on it, in February 2026) have pointed to a variety of structural problems. In particular, the Working Paper relies on rough estimate calculations

to arrive at  global royalty figures based on its estimate of over a billion copyright owners globally, which suggests that the amount paid to both authors and the entire population of copyright owners globally, may be so small as to not constitute compensating for their contributions to the production of an open-source work. For example, where an author owned only one track of an open-source work and OpenAI (assuming OpenAI had obtained a 10% royalty rate on its estimated $11 billion in annual revenue) would owe approximately $2.00 per work to the author of that work.

The no-opt-out design has also been severely criticised as being constitutionally suspect, commercial coercive and as arguably infringing on an author's moral rights as protected under Section 57 of Copyright Act, 1957. 8   The manner in which the fair dealing provision has been treated in relation to Section 52(1)(a) has also been described as a doctrinal shortcut and, therefore, an inadequate or incomplete analysis of whether TDM is a reproduction of copyright in the traditional sense of the word.

2.4 The Dispute Resolution Lacuna

An important shortcoming of the Working Paper is that it does not establish a specific system of dispute resolution for the mandatory licensing model that it proposes. The mandatory licensing model inherently creates at least five different kinds of disputes, namely: (i)

disputes about setting royalties between the CRCAT and AI developers; (ii) disputes between CRCAT and those claiming to be rights-holders over distribution; (iii) disputes about attribution and moral rights in AI derived from AI-generated works; (iv) disputes regarding infringement related to the initial act of using data to train AI; and (v) disputes regarding how CRCAT operates and should be held accountable. 10 Each type of dispute has its own set of legal and technical characteristics, which make standard litigation an inadequate and unfair

method of resolving these types of disputes.

III. Suitability of Mediation and Arbitration for Copyright–AI Disputes

3.1 The Global ADR Turn in IP Disputes

ADR now has become the most prominent mechanism for resolving IP disputes at an international level, which includes particularly in the copyright and digital content disputes. During 2025, WIPO (World Intellectual Property Organization) and the WIPO AMC (WIPO Arbitration and Mediation Centre) facilitated 1461 disputes involving the areas of intellectual property, innovation, and technology (a 70% increase from 2024). 11 71% of these mediations involved copyright and digital content disputes themselves, as well as a large number of disputes related to artificial intelligence and platforms for sharing content. In the same year, the WIPO AMC also experienced a growth of 87% in terms of the number of disputes co-administered with national offices focused on Intellectual Property and Copyright, indicating the increasing level of institutional interest in having embedded ADR

provisions at the regulatory level.

In India the ADR ecosystem for Intellectual Property is growing rapidly. The Mediation Act of 2023 is creating a structure for facilitating pre-litigation mediation and institutional mediation in civil and commercial disputes. Section 5 of the Act makes pre-litigation mediation mandatory for all commercial disputes; Section 18 requires that mediation must be completed no later than 120 days after it commences (extensible to 180 days if agreed) and Section 27 states that settlement agreements that have been authenticated will be treated as court decrees, this provision will be critical to providing the finality of ADR outcomes, especially in high-stakes disputes relating to IP licences.

There was a programmed that was launched in 2025 called the CGPDTM-WIPO AMC Boudhik Smadhan, which operates this statute particularly for IP disputes that contributes by providing a pathway to mediation which is voluntary in nature, and also cost effective. It is governed under Mediation Act of 2023, that is administered together by India’s IP

administration and the WIPO AMC. This can be adapted and put forward to proceed with the categories of disputes that seem to arise under the proposed DPIIT.

3.2 Mediation for Licensing Rate and Distribution Disputes

Disputes concerning royalty rates between CRCAT and AI developers are likely to be good candidates for mediation. They involve parties who are continuing to do business together, there are many types of technical and economic evidence, and all parties want predictable

outcomes to avoid costly litigation that could create a precedent for other disputes. Mediation is a non-adversarial process where parties will be able to present their economic models, usage data and royalty benchmarks in a confidential environment to address these types of

disputes given that the AI companies training data and projected revenue are considered commercial trade secrets.

The Mediation Act 2023 provides a clear fraud defence based on the Mediation Act’s confidentiality protections (the right to confidentiality of communication made between a lawyer and client during the mediation process) which protects parties to mediation from

having communications made during those proceedings used by a party against another party in a subsequent civil action. The work done by the WIPO AMC (World Intellectual Property Organizations Alternative Mediation Centre) with CMOs (Collective Management Organizations) or CMO-related matters or with CMOs dealing with tariff disputes is a good example of how the WIPO AMC can help in the resolution of modification of current licensing criteria and tariffs imposed between CMOs and their respective rightsholders as part of their digital copyright efforts in an evolving digital environment.

3.3 Arbitration for Infringement and Enforcement Disputes

Arbitration is the only way through which infringement disputes regarding artificial intelligence training data ingestion, and enforcement of a mandatory licence can be conclusively resolved, and thus, provides finality and enforceability. The Arbitration and Conciliation Act of 1996 is the backbone of both international and domestic commercial arbitration in India. Arbitral awards issued under the New York Convention are enforceable in over 170 countries, which is essential because many of the larger entities involved in artificial intelligence development are foreign and hold their assets outside of India.

For disputes involving copyright and artificial intelligence, a panel of specialist arbitrators, possessing expertise in copyright law, data science, and economic valuation, would be in a much better position than generalist judges to determine whether a particular AI practitioner’s

data ingestion exceeded the scope of a mandatory licence and the appropriate compensation for the infringement. There are currently WIPO AMC’s expert arbitration panels available which contain arbitrators with expertise in both AI and copyright; this panel will resolve

disputes more quickly through expeditious arbitration and will involve only a limited amount of evidence gathering compared to adversarial proceedings.

3.4 Hybrid Med-Arb for CRCAT Governance Disputes

A hybrid form of mediation and arbitration is generally suited to disputes of CRCAT’s governance, which include the challenges to the decisions of royalty allocation and distribution, method of accounting, and eligibility criteria. The mediation or arbitration process utilizes mediation first by allowing the parties involved to attempt to reach an agreeable result through an interest-based negotiation; if mediation cannot resolve the issue(s) within a specified period (e.g. 60 days), the mediator may be used to proceed with an arbitration award on the unresolved issues or a separate arbitrator appointed by the authority

will issue an awards. The mediation/arbitration process enables parties to reach probable consensual results. It also ensures that disputes are not left unresolved and with uncertainty affecting the entire licensing environment.

3.5 Expert Determination for Technical and Valuation Questions

Licensing disputes will frequently involve technical questions- for example, how many pieces from a dataset belong to an artist, or which royalty calculations are more accurate than others that are typically outside of the courts’ litigious abilities but lend themselves to expert

determination. Experts in the WIPO AMC may determine an answer to one defined technical or value-related question and will create a binding determination via their decision after reviewing the documents and/or evidence provided by either party.

As India has yet to develop an independent statute addressing disputes arising from licensing, parties can include an expert determination clause or term in an agreement which permits the parties to submit their questions to a qualified expert. A party may also appeal the expert's

decision through arbitration based on a claim of manifest error under the Arbitration and Conciliation Act of 1996, allowing party to ultimately have their questions addressed by a third party.

IV. Proposed ADR Framework for Implementation in India

4.1 Architecture

This paper puts forth a proposal for the creation of an Intellectual Property–Alternative Dispute Resolution (IP-ADR) Panel (through necessary amendments or rules) which would be created in the Copyright Office under the Copyright Act 1957. This new Panel would

work as an independent institutional ADR provider, with its own rules/procedures, roster of neutrals, and fee schedule and would provide services for the five (5) categories of disputes as outlined in Part II, with specific procedures for resolving each category of dispute. Disputes regarding the payment of royalties and distribution of royalties (Categories i. and ii.):

Disputes regarding payment of royalties and distribution of royalties would require all parties to engage in mandatory mediation not less than sixty (60) days prior to filing of any arbitration, followed by binding arbitration before a 3-member panel consisting of a copyright law expert, an economist and a data scientist. The ultimate decision rendered by the (3) arbitrators would be reasoned and made available in a redacted format to establish a body of precedent.

The mediation process will involve disputes regarding entitlement to the author's identity (i.e. settlement of evidence regarding moral and intellectual rights), where mediation will take place in 30 days; and where there is a need for expert testimony concerning technical issues

of authorship identity, and arbitration to resolve any legal issues concerning moral and intellectual rights. It is expressly provided that a right-holders; access to the courts will be available permittively for injunctive relief, in accordance with the United States Supreme

Court decision Yamini Manohar v. TKD Keerthi (2023), which establishes the continuing need to enforce IP rights through an immediate relief process notwithstanding a party’s requirement to pursue mediation.

Disputes regarding the use of training data for infringement (i.e. IP) will be submitted to arbitration under rules that mimic the World Intellectual Property Organization’s Expedited Arbitration rules; and awards shall be rendered within 90 days from the date of tribunal appointment to the date of award. The arbitration process shall allow for emergency

arbitration in cases in which a violation of an IN infringement occurs.

Disputes regarding the governance of the CRCAT shall proceed through a med-arb process wherein mediation will take place over 60 days and will automatically proceed to arbitration before a panel comprising members from the IP-ADR Panel.

4.2 Neutral Roster and Capacity Building

At present, India does not have enough qualified mediators and arbitrators with experience in copyright law combined with expertise in AI technology. The proposed framework should create a neutral roster of specialists, or neutrals, that will be created by a joint effort of the

Copyright Office, the Bar Council of India, NASSCOM and ADR institutions, such as the Mumbai Centre for International Arbitration (MCIA) and the Delhi International Arbitration Centre (DIAC). The creation of training programmes, similar to the WIPO ADR Young initiative, should also be undertaken in order to build capacity for the next generation. 20

4.3 Model ADR Clause

Each licence issued under the CRCAT framework should contain a standard ADR provision that provides for: (i) the IP/ADR panel as the exclusive venue for pre-litigation dispute resolution; (ii) the confidentiality of all ADR proceedings; (iii) the procedural rules for ADR;

(iv) New Delhi as the seat of arbitration; and (v) the parties; agreement to comply with any award or settlement agreement without having to pursue enforcement in court unless one party fails to comply within 30 days.

4.4 Online Dispute Resolution

Since rights holders under the national mandatory licensing scheme will have individual authors, composers, and journalists throughout the linguistic and geographical diversity of India as claimants, the IP-ADR Panel must have an online dispute resolution (ODR) system in place which is consistent with the ODR framework outlined by NITI Aayog. ODR will reduce the costs and barriers to access that might otherwise systematically exclude small rights holders from participating meaningfully in the dispute resolution process.

V. Challenges, Comparative Insights, and Recommendations

5.1 Comparative Insights

In the UK Copyright and AI Report (2026), a broad exception to TDM was rejected and unlike the EU AI Act and the DSM Directive, which both implement opt-out provisions with disputes concerning the ability of a party to opt-out referred to national courts or specialist IP tribunals, the UK TDM Reforms will rely on negotiated licenses through specialist courts and existing dispute resolution systems to settle any disputes concerning TDM. The TDM Exception under Article 30-4 of the Copyright Law of Japan has resulted in only a few disputes, in large part because it is premised on the broad spectrum of permissible action and is coupled with clearly defined statutory guidance on how to use TDM as a means of conflict avoidance (which is a lesson learned for India in the need for clarity in legislation to help avoid future conflicts).

The rise of copyright litigation in the United States has resulted in many cases, The New York Times Co. v. Microsoft Corp. and several author class actions against AI companies, as well as the extreme costs and delays due to not having an ADR in place, with some cases reaching two years of pretrial discovery, causing individual authors and smaller AI developers to be unable to afford the enormous legal fees involved.

5.2 Challenges and Recommendations

In order to successfully implement the proposed ADR framework, there are four challenges to overcome: (i) To create a large enough and adequately skilled neutral panel that can be created quickly; (ii) To permit the creation of public jurisprudence with respect to crucial legal standards (with that public jurisprudence being developed in a manner which balances confidentiality of ADR proceedings); (iii) To prevent powerful AI creators from being able to exploit the procedural difficulties created by arbitration statutes as they relate to smaller rights holders; and (iv) To connect the ADR framework to the courts; supervisory power in a way that does not create fragmented disputes.

Recommendations for addressing these challenges include: Amending to Copyright Act 1957 to create the establishment of an IP-ADR Panel and require ADR as a precondition to coming to court for license disputes; Providing that, through the enabling legislation for

CRCAT, redacted arbitration awards must be publicly available in order to develop public jurisprudence; Creating fee schedules based on means testing so as to ensure access by individual rightsholders; Creating a requirement that the ADR framework be reviewed within

3 years of its implementation (with follow-up reports to Parliament outlining rates of resolution, costs, and levels of satisfaction).

VI. Conclusion

The One Nation, One License, One Payment working paper issued by the DPIIT represents an important but controversial step towards a unified regulatory framework for AI training data in India. The mandatory blanket licensing structure will create disputes of tremendous technical and economic complexity, disputes that India's courts are fundamentally unprepared to resolve in a timely manner from a structural standpoint. The fact that the working paper

says nothing about how disputes will be resolved is not merely an oversight; it is a gaping hole that, if not addressed, could render the entire framework unworkable in practice.

Alternative dispute resolution (ADR) may provide the only truly fit-for-purpose mechanism for this problem. ADR includes: mediation under the Mediation Act, 2023; arbitration under the Arbitration and Conciliation Act, 1996; hybrid mediation-arbitration (med-arb); and expert determination. The Boudhik Smadhan program and co-

administration infrastructure between the WIPO and the Indian Government provide an available framework for building a dedicated IP-ADR Panel within the Copyright Office framework. WIPO’s 2025 stats show a 70% increase in the IP ADR caseload. The continued demand (and institutional viability) of this structure is statistically validated by the fact that copyright and digital content disputes represented the plurality of disputes filed.

The choices made around AI governance in India today will impact not only local creative sectors but also how India fits into the worldwide AI value chain. The arrangement for resolving disputes is not an afterthought; it is the factor that will determine whether any licensing scheme will work in practice as opposed to merely existing on paper. A well constructed regime for the alternative resolution of intellectual property (IP) disputes within the CRCAT framework will be India's most important and long-term contribution to the global governance of copyright in AI in the years 2025 and beyond

refrences

1. Department for Promotion of Industry and Internal Trade. Working Paper on Generative AI and Copyright – Part I: One Nation One License One Payment. Dec. 2025.

2. Kumar, Manish. "Summary: Working Paper on Generative AI and Copyright – Part I." ALG India Law Offices LLP, 22 Jan. 2026.

3. The Copyright Act, 1957, No. 14, Acts of Parliament, 1957 (India), s. 14.

4. Barooah, Swaraj Paul. "Bulldozer Justice Comes to the Digital Sphere: Looking at the GenAI-Copyright Report." SpicyIP, 15 Dec. 2025.

5. The Copyright Act, 1957, No. 14, Acts of Parliament, 1957 (India), s. 14.

6. Kaushik, Shivam. "One Nation, One License, One Big Shortcut: Doctrinal Stagnation in the DPIIT AI Working Paper." SpicyIP, 20 Dec. 2025.

7. Barooah, Swaraj Paul, et al. "Comments to the DPIIT on Their GenAI – Copyright Working Paper 1." SpicyIP, 6 Feb. 2026.

8. Byadwal, Yogesh. "DPIIT Working Report: Lawful Access While Ignoring Copyright Law." SpicyIP, 9 Jan. 2026.

9. The Copyright Act, 1957, s. 52(1)(a).

10. Aggarwal, Ambika. "One Nation, Forced Licenses, Multiple Payments: (Un)Balancing AI Innovation & Copyright." SpicyIP, 14 Jan. 2026.

11. WIPO Arbitration and Mediation Center. "WIPO ADR Highlights 2025.; WIPO, 2026.

12. European Commission. "WIPO ADR Highlights 2025." IP Helpdesk, 10 Mar. 2026

13. The Mediation Act, 2023, No. 32, Acts of Parliament, 2023 (India), ss. 5, 18, 27.

14. Verma, Bhavya. "India's Mediation Act of 2023." Asia IP, 8 Dec. 2023.

15. IP India – O/o CGPDTM and WIPO AMC. "Boudhik Smadhan : IP Mediation Programme Guidelines." Government of India, 2025.

16. “What the New Mediation Law Means for IP Disputes in India." Chambers and Partners, 2024.

17. WIPO Arbitration and Mediation Center. "Caseload Summary 2016–2025." WIPO, 2026

18. Toscano, Leandro, et al. "ADR for Copyright and Content Disputes in the Digital Environment." WIPOD Arbitration and Mediation Matters.

19. Supreme Court Ruling on Impact of Mediation Delays on IP Injunctions." IAM Media, 10 Dec. 2025.

20. Where Mediation Meets IP: A Young Lawyer's View of India's Legal Shift." Bar and Bench, June 2026

21. Report on Copyright and Artificial Intelligence. UK Government, 2026.

22. Agrawal, Akshat. Part 1: Copyright is the Wrong Answer.SpicyIP, 7 Jan. 2026.

23. Sudheendra, Vishno. ;Copyright Maximalism by Design? Rethinking DPIIT';s Licensing-Centric Approach to AI Training; SpicyIP, 21 Jan. 2026

24. Can Mediation Be Used to Resolve Intellectual Property Disputes in India?; Bar and Bench, 27 Jun. 2024.