Algorithm Licensing Claims in India

1. Meaning and Scope

Algorithm Licensing Claims arise when an algorithm, AI model, software system, source code, database, or algorithm-enabled service is supplied under a licence and a dispute develops concerning the right to use, modify, reproduce, distribute, commercialise, sublicense, access, or continue using the technology.

Indian law does not recognise a standalone statutory cause of action called an “algorithm licensing claim.” Such disputes generally arise through a combination of:

contract law;

copyright law;

software licensing;

trade-secret/confidentiality principles;

patent law where applicable;

information-technology law;

data-protection law;

consumer law in appropriate cases;

arbitration;

injunction and specific-relief principles.

Algorithm licensing disputes can arise between:

AI developer and enterprise customer;

software company and government department;

algorithm owner and distributor;

technology licensor and licensee;

platform and API customer;

employer and technology vendor;

research institution and commercial licensee;

software developer and former employee;

two competing technology companies.

2. What Is an Algorithm Licence?

An algorithm licence is essentially a contractual permission allowing another party to use specified technological rights subject to agreed conditions.

The licence may cover:

source code;

object code;

executable software;

mathematical models;

machine-learning models;

model weights;

APIs;

documentation;

datasets;

training materials;

improvements;

updates;

derivative works;

proprietary methods.

The licence can be:

Exclusive

Only the licensee receives the specified rights.

Non-exclusive

The licensor can license the same technology to others.

Sole

The licensor may retain some rights but agrees not to license it to additional third parties.

3. Algorithm Licensing Is Usually a Contractual Relationship

The first legal document to examine is normally the licence agreement.

Important clauses include:

scope of licence;

permitted users;

permitted territories;

duration;

permitted purposes;

modification rights;

copying rights;

sublicensing;

assignment;

API access;

reverse engineering;

benchmarking;

training of AI systems;

commercialisation;

ownership of improvements;

confidentiality;

security;

audit rights;

fees;

royalties;

termination;

post-termination obligations.

A dispute frequently arises because the licensee believes:

“I paid for the software, therefore I own it.”

That is generally incorrect.

Ownership and licensed use are legally different concepts.

4. Ownership vs Licence

Suppose Company A develops an algorithm and licenses it to Company B.

Company B may receive the right to:

install it;

run it;

access an API;

use it internally.

But Company A may retain:

copyright;

source-code ownership;

model ownership;

patent rights;

trade secrets;

rights to commercialise the technology elsewhere.

Therefore:

Payment for software does not automatically transfer intellectual-property ownership.

The agreement must be examined to determine what rights actually transferred.

5. Copyright Protection of Algorithms

This is one of the most important legal questions.

Under Indian copyright law, an algorithm as a mere mathematical idea or method of operation is not necessarily protected as copyright simply because it is described as an algorithm.

However, particular expressions implementing the algorithm may receive protection.

For example:

Abstract algorithm

“Sort records according to numerical value.”

versus

Computer program implementing the algorithm

Specific source code and object code implementing that process.

The second may attract copyright protection under the applicable provisions of the Copyright Act, 1957.

6. Section 13 and Computer Programs

The Copyright Act, 1957 recognises copyright protection for specified categories of works, including computer programmes.

The legal distinction between an underlying idea and its expression is therefore important.

A licensing claim should identify precisely what is being licensed:

algorithmic concept;

source code;

object code;

documentation;

database;

model weights;

interface;

proprietary implementation.

7. R.G. Anand v. Deluxe Films

R.G. Anand v. Deluxe Films, (1978) 4 SCC 118

This is a foundational Supreme Court authority concerning the idea-expression distinction in copyright law.

Algorithm licensing relevance

A licensor cannot necessarily claim copyright over a mere abstract idea or concept.

However, copying a protected expression implementing the concept can create infringement issues.

For algorithms, the question can therefore become:

Is the alleged copying of an unprotected algorithmic idea, or of protected expression such as source code?

That distinction can substantially affect a licensing dispute.

8. Eastern Book Company v. D.B. Modak

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

The Supreme Court examined originality under copyright law and emphasised the requirement of sufficient intellectual creation rather than mere labour or mechanical effort.

Algorithm relevance

Where a licensor claims copyright over:

software documentation;

code;

compilations;

technical material;

the claimant must establish the relevant copyright interest.

The fact that a company spent substantial money developing an algorithm does not automatically mean that every underlying mathematical idea becomes copyright-protected.

9. Engineering Analysis Centre of Excellence Pvt. Ltd. v. CIT

Engineering Analysis Centre of Excellence Pvt. Ltd. v. Commissioner of Income Tax, (2021) 432 ITR 471 (SC)

This is an extremely important Supreme Court judgment for software licensing.

The Supreme Court considered payments relating to software and the distinction between:

transfer of copyright rights; and

purchase of a copyrighted article/product.

Algorithm licensing relevance

The judgment demonstrates why the precise nature of the rights granted under a software arrangement matters.

A licence allowing a customer merely to use software is not necessarily equivalent to a transfer of copyright ownership.

This distinction can affect:

tax treatment;

contractual interpretation;

intellectual-property rights;

rights of reproduction;

commercial exploitation.

10. Tata Consultancy Services v. State of Andhra Pradesh

Tata Consultancy Services v. State of Andhra Pradesh, (2005) 1 SCC 308

The Supreme Court considered the legal character of software in the context of taxation.

Algorithm licensing relevance

The judgment is useful in understanding that software can have both:

intellectual-property characteristics; and

commercial/product characteristics.

In licensing disputes, the court must therefore examine the actual transaction rather than rely solely on labels such as:

“software licence,”
“technology service,” or
“algorithm subscription.”

11. Microsoft Corporation v. Yogesh Papat

Microsoft Corporation v. Yogesh Papat, 2005 (30) PTC 245 (Delhi)

The Delhi High Court dealt with unauthorised use of copyrighted software.

Algorithm licensing relevance

It illustrates the importance of obtaining lawful authorisation for software use.

A licensee who exceeds the permitted licence may potentially face:

infringement proceedings;

injunction;

damages/account of profits;

termination;

contractual claims.

12. Autodesk, Inc. v. Prakash Manghat

Autodesk, Inc. v. Prakash Manghat, 2008 (36) PTC 354 (Delhi)

The case concerns protection of software and unauthorised use.

Relevance

Software licensing disputes can involve not only contractual breach but also intellectual-property infringement.

If the license permits use only on a certain number of machines, for example, use beyond the permitted scope may generate both contractual and IP consequences depending on the licence terms and applicable law.

13. Bharat Broadband Network Ltd. v. United Telecoms Ltd.

Bharat Broadband Network Ltd. v. United Telecoms Ltd., (2019) 5 SCC 755

This Supreme Court decision is principally important in the context of arbitration and appointment of arbitrators.

Algorithm licensing relevance

Technology licence agreements frequently contain arbitration clauses.

When an algorithm licensing dispute arises, questions can include:

whether arbitration is mandatory;

whether the arbitration agreement is valid;

whether the dispute falls within the clause;

whether the arbitrator is legally eligible;

whether interim relief is required.

Thus, the dispute-resolution clause can be as important as the IP provisions.

14. Percept D'Mark (India) Pvt. Ltd. v. Zaheer Khan

Percept D'Mark (India) Pvt. Ltd. v. Zaheer Khan, (2006) 4 SCC 227

The Supreme Court examined contractual restrictions and injunction-related issues.

Algorithm licensing relevance

Technology agreements often contain post-termination restrictions.

Examples:

no continued use;

no competing deployment;

no disclosure;

no sublicensing;

no reverse engineering.

But contractual restrictions must be examined under applicable Indian law. A licensor cannot assume that every restrictive clause will automatically be enforceable.

15. Indian Contract Act, 1872

The Indian Contract Act, 1872 forms the basic contractual foundation for algorithm licensing disputes.

Important provisions may include:

Section 10

Requirements for valid contracts.

Section 17

Fraud.

Section 18

Misrepresentation.

Section 19

Voidability in certain cases involving absence of free consent.

Section 23

Agreements opposed to public policy or unlawful objects.

Section 27

Agreements in restraint of trade.

Section 37

Obligations of parties to contracts.

Section 39

Effect of refusal to perform promise wholly.

Section 55

Failure to perform at a fixed time where time is essential.

Section 73

Compensation for breach.

Section 74

Compensation where a penalty or stipulated sum is specified.

These provisions can become relevant depending upon the licence agreement.

16. Breach of Algorithm Licence

A licensee may breach the agreement by:

using the algorithm beyond the permitted purpose;

exceeding user limits;

sublicensing without permission;

sharing source code;

reverse engineering where prohibited;

allowing unauthorised third-party access;

copying the software;

modifying protected code contrary to the licence;

using the model after termination;

circumventing technical restrictions;

using the technology outside the permitted territory.

The licensor may seek contractual and/or intellectual-property remedies depending upon the circumstances.

17. Licensor Breach

The licensor can also breach the licence.

Examples include:

failure to provide agreed functionality;

failure to provide updates;

failure to maintain API availability;

supplying defective code;

failure to provide promised documentation;

failure to provide agreed support;

licensing technology without authority;

third-party IP infringement;

failure to meet security obligations.

A licensee may therefore have claims for:

damages;

specific relief where available;

termination;

refund;

indemnification;

arbitration;

other contractual remedies.

18. Misrepresentation in Algorithm Licensing

Technology vendors may represent that an algorithm is:

99% accurate;

bias-free;

secure;

proprietary;

compliant with law;

trained on licensed data;

suitable for a particular industry;

capable of a particular task.

If those representations are materially false, contractual and statutory remedies may arise depending upon the circumstances.

Example

A vendor states:

“Our algorithm has been trained exclusively on licensed datasets.”

The customer later discovers that the vendor incorporated unauthorised third-party material.

Possible issues include:

misrepresentation;

breach of warranty;

indemnity;

IP infringement;

termination.

19. Algorithm Licensing and Indemnity

Sophisticated licensing agreements often contain IP indemnities.

For example:

Licensor agrees to indemnify the customer against third-party copyright claims arising from authorised use of the licensed software.

Such clauses should be examined for:

scope;

exclusions;

notification requirements;

control of litigation;

settlement rights;

caps;

consequential-loss exclusions.

An indemnity clause can substantially determine who ultimately bears the cost of an infringement claim.

20. Algorithm Licensing and Confidentiality

Algorithms may contain valuable trade secrets.

Examples include:

model architecture;

optimisation methods;

source code;

proprietary parameters;

feature engineering;

training methodologies;

deployment techniques.

A licensing agreement may therefore contain strict confidentiality provisions.

A licensee that discloses confidential algorithmic information may face:

contractual damages;

injunction;

confidentiality claims;

IP-related proceedings where applicable.

21. Reverse Engineering

Many algorithm licences restrict reverse engineering.

A licensee may attempt to:

decompile software;

extract source code;

reproduce architecture;

derive model weights;

reconstruct proprietary functionality.

Whether such conduct is lawful depends upon:

the licence terms;

applicable copyright law;

statutory exceptions;

purpose of the reverse engineering;

competition-law considerations where relevant.

The contract should therefore define reverse-engineering rights clearly.

22. Algorithm Licensing and AI Model Weights

Modern AI licences can be more complicated than conventional software licences.

The licensor may license:

model weights;

API access;

inference rights;

fine-tuning rights;

deployment rights.

For example:

Customer may use the model through an API but may not download model weights.

Alternatively:

Customer may download the model but may not redistribute or commercialise derivative models.

These distinctions should be expressly documented.

23. Fine-Tuning and Derivative Models

A major modern licensing issue is:

Who owns a model modified or fine-tuned by the licensee?

Possible contractual models include:

Licensor ownership

All improvements remain with the original developer.

Licensee ownership

The customer owns its fine-tuned version.

Shared rights

Each party retains specified rights.

Field-of-use arrangement

Licensee owns the improvement but can only deploy it in a specified industry.

Without clear drafting, significant disputes can arise.

24. Algorithm Licensing and Data

An algorithm licence does not necessarily grant a licence to use every dataset processed by the algorithm.

There may be separate rights in:

personal data;

confidential business information;

copyrighted databases;

customer data;

government data;

third-party datasets.

Therefore:

Licence to use an algorithm ≠ licence to use all data processed by the algorithm.

This distinction is particularly important for AI systems.

25. Algorithm Licensing and Privacy

Where licensed software processes personal data, the parties should separately examine applicable data-protection obligations.

The contract should identify:

who determines the purpose of processing;

who processes the data;

security responsibilities;

breach notification;

retention;

deletion;

subcontractors;

cross-border transfers;

data-subject rights.

A technology licence cannot automatically override mandatory data-protection law.

26. Open-Source Algorithm Licensing

Open-source software creates a different licensing environment.

An algorithm may be distributed under licences such as:

permissive licences;

copyleft licences;

source-disclosure licences;

custom open-source-compatible terms.

A company incorporating open-source code must understand its obligations regarding:

attribution;

notices;

source-code availability;

modification;

redistribution;

licence compatibility.

Failure to comply can create both contractual and copyright-related disputes.

27. Algorithm Licensing and Patent Rights

An algorithm as a mathematical method is not automatically patentable in India.

However, a computer-related invention may raise patent questions depending upon the actual technical contribution and applicable patent law.

The Patents Act, 1970, particularly the restrictions concerning computer programmes per se and mathematical methods, must therefore be considered.

A licensing agreement should clearly state whether patent rights are included.

28. Copyright Licence vs Patent Licence

These are separate rights.

A software licence may authorise:

copying and using software.

But that does not necessarily grant:

patent rights covering a technical invention implemented by the software.

Similarly, a patent licence does not necessarily give permission to reproduce copyrighted source code.

A comprehensive technology agreement should therefore specify the relevant IP categories.

29. Government Algorithm Licensing

Government departments increasingly procure:

AI systems;

analytics platforms;

predictive tools;

automated decision systems.

Government algorithm licences raise additional issues involving:

public procurement;

transparency;

audit rights;

public-sector data;

security;

confidentiality;

intellectual property;

government access;

termination;

vendor lock-in.

The agreement should ideally preserve the government's ability to audit and verify the system where legally appropriate.

30. Tata Cellular and Algorithm Procurement

Tata Cellular v. Union of India, (1994) 6 SCC 651

Government procurement is subject to judicial review for recognised grounds such as:

arbitrariness;

irrationality;

procedural impropriety;

mala fides.

Algorithm licensing relevance

If a government selects a proprietary AI vendor through an unfair procurement process, competing vendors may challenge the procurement decision.

Therefore:

Technical sophistication does not eliminate public-procurement law.

31. Engineering Analysis and Licence Interpretation

Engineering Analysis Centre of Excellence is particularly valuable because it demonstrates the importance of determining exactly what rights a software transaction grants.

For algorithm licensing, lawyers should ask:

Is the customer receiving copyright rights?

Is the customer merely receiving a copy?

Is the customer receiving a right to use?

Can the customer reproduce the software?

Can the customer modify it?

Can it sublicense?

Can it commercially distribute it?

The answers can materially alter the legal character of the transaction.

32. Remedies for Algorithm Licensing Breach

Potential remedies include:

Damages

Compensation for proven contractual loss.

Injunction

Preventing unauthorised use, disclosure or distribution.

Specific performance

Available in appropriate circumstances under the Specific Relief Act, subject to statutory limitations.

Account of profits

Potentially relevant to certain IP infringement claims.

Termination

Where the contract permits termination for material breach.

Arbitration

Where the licence contains a valid arbitration agreement.

Declaratory relief

Clarification of contractual/IP rights in appropriate proceedings.

33. Important Case Laws at a Glance

34. Core Questions in an Algorithm Licensing Dispute

A court or arbitral tribunal would typically need to determine:

1. What exactly was licensed?

Algorithm? Software? Source code? API? Model weights?

2. Who owns the underlying IP?

Licensor, licensee or third party?

3. What rights were granted?

Use? Copy? Modify? Distribute? Sublicense?

4. Was the licence exceeded?

Did the licensee go beyond the contractual scope?

5. Was the licensor authorised to license the technology?

This becomes critical where third-party code or data is incorporated.

6. Were representations accurate?

Was the technology represented accurately?

7. Was there infringement?

Did either party infringe copyright, patent or other protected rights?

8. What happens upon termination?

Does the customer have to:

delete software;

destroy copies;

cease API access;

return materials;

delete model weights?

9. Which dispute mechanism applies?

Court litigation or arbitration?

35. Practical Example

Suppose Company A develops an AI fraud-detection algorithm.

It grants Company B a five-year licence allowing:

internal use;

deployment on 100 servers;

no sublicensing;

no source-code disclosure.

Company B subsequently:

installs it on 500 servers;

gives access to its subsidiary;

modifies the source code;

sells the modified algorithm to customers.

Company A could potentially allege:

breach of licence scope;

unauthorised reproduction;

unauthorised modification;

unauthorised sublicensing;

copyright infringement;

breach of confidentiality;

contractual damages.

The precise claims depend upon the licence language and applicable law.

36. Common Defences by Licensees

A licensee may argue:

the conduct falls within the licence;

the licence permits modification;

the agreement transferred ownership;

the licensor consented to the use;

the alleged material is not copyright-protected;

the restriction is unenforceable;

the licensor lacks ownership;

the claimant has not established infringement;

the alleged breach caused no recoverable loss;

contractual limitations apply;

the licensor waived strict compliance.

37. Drafting an Effective Algorithm Licence

A robust agreement should clearly define:

Intellectual property

Exactly what constitutes the licensed technology.

Scope

Who may use it and for what purposes.

Territory

Where it may be deployed.

Duration

When rights begin and end.

Modification

Whether the licensee can modify or fine-tune the system.

Derivative works

Who owns improvements.

Data

Who owns and controls input and output data.

Security

Minimum technical safeguards.

Audit

Licensor's ability to verify compliance.

AI-specific provisions

model training;

fine-tuning;

benchmarking;

reverse engineering;

model extraction;

prompt/output rights;

synthetic data;

redistribution.

Termination

What happens to software, copies, models and data after termination.

38. Key Legal Distinction: Algorithm vs Implementation

One of the most important principles in algorithm licensing is:

The legal protection of an algorithmic idea is not necessarily the same as the legal protection of the software implementation embodying it.

Therefore, a claimant should avoid making an overly broad assertion such as:

“I own the algorithm, therefore nobody can use the concept.”

The more precise question is:

What legally protected right exists, and what contractual licence governs that right?

39. Conclusion

Algorithm Licensing Claims in India are primarily a combination of contract and intellectual-property disputes, supplemented where appropriate by confidentiality, data protection, patent, arbitration and procurement law.

The most important authorities include:

R.G. Anand v. Deluxe Films — idea versus expression;

Eastern Book Company v. D.B. Modak — originality;

Tata Consultancy Services v. State of Andhra Pradesh — software as a legally significant commercial subject;

Engineering Analysis Centre of Excellence v. CIT — software licensing and copyright rights;

Microsoft Corporation v. Yogesh Papat — unauthorised software use;

Autodesk v. Prakash Manghat — software protection;

Percept D'Mark v. Zaheer Khan — contractual restrictions;

Bharat Broadband Network v. United Telecoms — arbitration;

Tata Cellular v. Union of India — government procurement.

The central legal formula is:

Protected Technology + Valid Licence + Defined Scope + Contractual/IP Rights + Alleged Overuse or Breach = Algorithm Licensing Dispute.

For modern AI licences, the most important drafting issues are model ownership, source-code access, API rights, model weights, training and fine-tuning rights, derivative models, data rights, sublicensing, reverse engineering, confidentiality, security, audit rights, indemnification, termination and post-termination deletion.