Intellectual Property Issues with AI

Artificial Intelligence (AI) is transforming every sector of society—from education and healthcare to law, entertainment, and business. However, as AI systems become more advanced and capable of generating content, designs, inventions, and databases, complex intellectual property (IP) challenges have emerged.

India, like many other jurisdictions, is grappling with these questions:

• Can AI-generated works enjoy copyright protection?
• How should trade secrets apply to AI models and algorithms?
• What is the scope of database protection when AI uses large volumes of information for training?
• Can AI-related inventions be patented under the Indian Patents Act, 1970?
• How do copyright laws respond to the training of large AI models on copyrighted datasets?

This article provides a comprehensive analysis of IP issues in relation to AI in India, covering these five important aspects in detail.

Can AI-Generated Works Be Copyrighted in India?

Copyright Law in India: Human-Centric Framework

Copyright in India is governed by the Copyright Act, 1957. The Act grants protection to original literary, artistic, musical, and dramatic works, as well as cinematograph films and sound recordings. Importantly, copyright arises from originality and authorship, where the author is traditionally a natural person.

The central issue is: if AI generates a poem, music track, design, or painting without human involvement, who is the “author”?

Current Position in India

• The Copyright Act does not recognise AI as an author or legal person.
• Section 2(d) defines “author” differently for each category of work, but in every case, the author is a human being or a legal entity (like a company) represented by humans.
• For computer-generated works, the author is deemed to be the person who causes the work to be created.

Thus, if an AI tool creates an artwork, the copyright would likely vest in the human programmer, user, or entity directing the AI, not the AI itself.

Practical Implications

• Ownership disputes may arise: Was it the developer, the company using the AI, or the end-user who “caused” the work to be made?
• Moral rights (like the right to attribution) are challenging to apply when a machine produces the content.
• Enforcement of rights: Without clear recognition of AI authorship, enforcing copyright in AI-generated works can be legally uncertain.

Possible Reforms

India may eventually need to amend its copyright law to:

• Define the role of AI in authorship, and
• Provide clarity on ownership of purely AI-generated works.

For now, Indian law does not protect AI-generated works as independent authorships—copyright flows back to human involvement.

Trade Secrets and AI Models

Why Trade Secrets Matter for AI

AI development involves algorithms, source codes, datasets, and model architectures. Companies often protect these through trade secrets, especially where patenting may be difficult or disclosure-heavy.

For example, large language models (LLMs) like ChatGPT or proprietary image recognition tools depend on unique architectures and training datasets. Disclosing these may destroy their commercial value.

Trade Secret Protection in India

India does not have a specific trade secrets legislation, but protection is available under:

• Contract law (confidentiality agreements, NDAs),
• Equity principles (common law action against breach of confidence),
• Employment law (restricting misuse of confidential information).

Challenges with AI Models

1. Reverse Engineering: AI models can often be reverse-engineered through “model extraction attacks,” raising difficulties in keeping them secret.
2. Employee Mobility: Developers or engineers moving to competitors may expose confidential model architecture or training techniques.
3. Training Data Transparency: If training data includes sensitive or proprietary information, companies risk exposure of trade secrets.

Best Practices

• Strong contractual protections with employees and collaborators.
• Clear documentation of trade secret policies.
• Technical safeguards such as encryption, access controls, and secure APIs.

In India, trade secrets remain the most practical way of protecting AI models, though lack of a dedicated statute may create enforcement challenges.

AI and Database Protection

The Role of Databases in AI

AI models, particularly machine learning systems, rely heavily on databases. These databases may include:

• Text corpora,
• Medical records,
• Financial data,
• Images, videos, or sounds.

Databases can themselves be valuable IP assets, and protecting them is a key concern.

Copyright in Databases under Indian Law

Under the Copyright Act, compilations, including databases, are protected as “literary works” provided they meet the test of originality. This means the database must involve selection or arrangement requiring skill, labour, and judgement, not mere mechanical collection of data.

Challenges in the AI Context

• Large-Scale Scraping: AI training often relies on web-scraped databases. Many such databases lack copyright originality, but some may be protected.
• Ownership Disputes: Multiple contributors may create uncertainty in database ownership.
• Fair Use/Exceptions: Section 52 of the Copyright Act allows limited use of works for research or private study, but large-scale AI training may not fit neatly into these exceptions.

Database Rights in Other Jurisdictions

In the EU, databases enjoy a sui generis right protecting substantial investment in obtaining, verifying, or presenting data. India currently has no such parallel.

Patentability of AI Inventions under the Patents Act, 1970

Patent Law in India

The Patents Act, 1970 protects inventions that are:

• Novel,
• Involve an inventive step, and
• Are capable of industrial applcation.

However, Section 3 of the Act lists several non-patentable inventions—including mathematical methods, business methods, and computer programs per se.

Can AI Itself Be an Inventor?

Globally, debates have arisen around whether AI can be named as an inventor. For example, in the “DABUS case,” some jurisdictions rejected AI inventorship, while others considered limited recognition.

In India:

• The Patents Act requires the inventor to be a natural person.
• Patent applications must be filed by human inventors or their assignees.
• AI systems cannot be treated as “inventors” or “persons” under Indian law.

Patentability of AI-Related Inventions

While AI itself cannot be the inventor, AI-related inventions may be patentable if they meet the statutory criteria. Examples include:

• Novel algorithms combined with hardware applications (beyond “computer programs per se”).
• AI-based drug discovery methods with specific industrial applications.
• AI-driven control systems in manufacturing, robotics, or telecommunications.

Key Challenges

• Section 3(k) of the Act excludes algorithms and computer programs per se, unless they show technical application or produce a “technical effect.”
• Obviousness: AI models often involve training data and optimisation, which may be considered routine rather than inventive.
• Disclosure Requirements: Patent applications must sufficiently disclose the invention. For AI, this can be problematic due to “black box” decision-making.

In India, patentability of AI inventions remains limited but possible when the invention demonstrates technical effect and industrial applicability, not merely abstract algorithms.

Copyright Challenges in Training Large AI Models

AI Training and Copyright Concerns

Training modern AI requires massive datasets that include copyrighted works—books, images, music, films, and journalistic content. This raises pressing copyright issues:

1. Reproduction Right: Training AI involves copying works into datasets, potentially infringing copyright.
2. Derivative Works: AI outputs may sometimes closely resemble training data, raising concerns of unauthorised adaptation.
3. Moral Rights: Indian law protects authors’ rights of attribution and integrity. If AI outputs distort or reproduce an author’s work, moral rights may be violated.

Indian Copyright Act Framework

• Section 14 grants copyright owners the exclusive right to reproduce, distribute, and adapt their works.
• Section 52 provides exceptions like fair dealing for research, criticism, or review. However, these are narrowly interpreted in India.
• Large-scale AI training may not qualify as “fair dealing” because it involves massive copying for commercial use, not limited private study.

International Developments

Other jurisdictions are considering “text and data mining (TDM) exceptions” to copyright. India does not yet have a TDM-specific framework.

Indian Challenges

• Uncertainty: AI companies in India face legal risks when training models on copyrighted material without licences.
• Licensing Complexity: Obtaining permissions for millions of works is practically unfeasible.
• Enforcement: Authors and publishers may face difficulties detecting and proving infringement by AI outputs.

Potential Solutions

• India could introduce a TDM exception with safeguards for rights holders.
• Collective licensing mechanisms may allow AI developers to access works legally.
• Clearer guidelines on fair dealing for AI research could strike a balance between innovation and copyright protection.

Conclusion

Artificial Intelligence brings enormous opportunities but also challenges the traditional contours of intellectual property law in India.

• Copyright: AI cannot be recognised as an author; protection flows back to the human directing the AI.
• Trade Secrets: Crucial for safeguarding AI models and algorithms, though enforcement is based on contracts and common law.
• Databases: Protected as compilations under copyright, but lack of a sui generis right leaves gaps.
• Patents: AI inventions may be patentable only if they show technical effect and industrial application, though AI itself cannot be an inventor.
• Copyright in AI Training: India faces uncertainty on whether mass copying for AI training qualifies as fair dealing or infringement.

For India, the key challenge is balancing innovation with protection of rights. As AI continues to evolve, legislative and judicial clarifications will be essential to ensure that intellectual property law keeps pace with technological advancements.