Stealing Ideas: Intersection Of AI-Generated Art And Copyright Law

Introduction

AI-generated art learning models are trained on massive datasets of art has brought about a set of legal challenges at the intersection of technology and copyright. Though the Indian Copyright Act, 1957, protects original art, this act is not specifically clear on protecting AI-generated art and nor does it provide clarity on the authorship of a piece of art if it is created by a machine. Since AI methods frequently learn from available public data without formal licenses, unauthorized copying of copyrighted material represents possible copyright infringement. While copyright shall not subsist in ideas or artistic styles as such, if AI-generated works for commercial purposes were to copy unique styles, or deceive the public, there will be a chance that the principles of passing off or unfair competition apply. This changing landscape raises questions about how robust existing legal frameworks will prove to be in responding to ownership, originality, and misuse in the era of AI-generated creativity.

What is AI-generated Artwork?

AI-generated art is artwork created by artificial intelligence algorithms trained on large datasets of artworks. Machine learning algorithms, particularly deep neural networks, examine patterns, style, and features to create new images. The process involves selecting datasets, training the AI, generating images based on learned patterns, and refining outputs through techniques like style transfer. Popular tools such as Deep Dream, StyleGAN, and CycleGAN exemplify computational creativity blended with artistic expression.

Current Position of Copyright Law in India

Section 13 – Works in which Copyright Subsists

Section 13(1) grants copyright protection to original literary, dramatic, musical, and artistic works; cinematograph films; and sound recordings.

Section 13(2) imposes territorial conditions: Published works must be published in India or authored by an Indian citizen at the time of publication/death. Unpublished works require the author to be domiciled in India or an Indian citizen for a substantial period. Architectural works must be situated in India.

Section 17 – First Ownership of Copyright

The author is generally the first owner, except: Works created under contract for newspapers/magazines (the publisher owns unless agreed otherwise). Works made for valuable consideration at someone's request (requester owns). Works created by employees under contract (employer owns unless agreed otherwise). Public speeches, government worksor works under the direction of public bodies have specific ownership rules.

Authorship vs. Ownership

Authorship: The person who creates the work through original intellectual effort. The author is always acknowledged as the creator, even if not the owner.

Ownership: The person or entity with legal rights to exploit the work. Ownership can be assigned or transferred.

Copyright in AI-Generated Work

The emergence of AI software that creates original creative content, such as images, music, and writing, poses a threat to traditional copyright law, which generally protects original human-created works. Because AI results in outputs without explicit human authorship, existing laws generally fail to recognize such works as copyrightable. Whereas users who submit prompts assert ownership, their contribution typically is not quite creative enough to be considered authorship, and the developers of AI who construct the tools are also unable to assert rights as they do not directly create the final piece of work. This is a legal grey area that would place many AI-made works outside copyright. Moreover, there are ethical implications where artists are concerned that AI will devalue human creativity and livelihoods, particularly when AI is trained on current work without permission, issues of fairness and originality come up. Generally, AI content challenges core copyright principles of creation and ownership, with the law still developing to tackle such complexities.

Copyright Infringement by AI

In India, the Copyright Act of 1957 doesn't really tackle the issue of AI-generated content head-on, which leaves a lot of questions about who owns what and what counts as original work. According to Section 2(d)(vi), the person who gets a computer-generated work made is considered the author, but it's still a bit murky whether AI fits the bill as a "computer" in this context. Then there's Section 2 (ffc), which describes a "computer program" as a set of instructions for carrying out tasks—this could eventually include AI prompts, but it also raises some red flags about potential misuse. While the Act doesn't define originality, it's a key factor for protection under Section 13(a). Indian courts look at concepts like Sweat of the Brow, Modicum of Creativity, and Skill and Judgment to determine originality, and AI-generated content might only qualify if there's enough human input involved. Section 17 talks about who the 'first owner' is for a work created for payment, which could back up a prompt-giver who pays to use an AI tool, but ownership is still a grey area, especially with minimal human involvement. Liability is a complicated matter. When AI generates content that violates rights held by others, it becomes unclear who ought to be held accountable: the user, the developer, or the AI itself, particularly because present Indian law does not consider AI to be a legal person. The ambiguity really highlights the imperative for India to modernize its intellectual property legislations Adapting these laws is crucial for aligning with international standards and clearly defining key concepts like authorship, ownership, originality, and liability in relation to works generated by AI.

Copyright Infringement by AI under Fair Dealing

In the ANI v. OpenAI case currently unfolding, the Delhi High Court is examining whether the use of proprietary content to train a large language model (LLM) and its reproduction in AI-generated responses amounts to copyright infringement or if it can be justified under the concept of fair dealing. Although Indian copyright law doesn't clearly adopt the U.S. idea of transformative use, courts might still consider whether an AI's functionality is more about analysis than just expression, and if it's aimed at a new objective—like providing commentary, conducting research, or offering criticism—potentially fitting within the fair use framework. To be considered fair dealing, AI-generated material typically has to fulfil three main criteria:

(A) Purpose – the use should be transformative, educational, critical, or focused on research.

(B) Extent of Use – keeping it minimal or utilizing content from the public domain generally supports fair use.

That said, issues can arise in cases of regurgitation (where protected material is copied word-for-word) or hallucination (where false or misleading information is created), which can make fair use claims more complicated.

AI Generated Art under Copyright: Idea vs Expression Dichotomy

The concept of the idea-expression dichotomy is very important when addressing the copyright concerns of AI-generated works. It aids in distinguishing between unprotected ideas and protectable expressions. Generative AI models are learned through big datasets, which sometimes comprise copyrighted materials, but they predominantly learn general patterns, styles, and concepts—these are types of ideas, which are not protected by copyright law. According to Indian copyright law, under Section 14 of the Copyright Act, 1957, only the unique expression of an idea and not the idea itself is protected. So, if AI generates new work by interpreting and rearranging learned patterns without outright copying definite aspects such as dialogue, character names, or plot, it generally doesn't violate copyright. But if the output of the AI is remarkably close to the work's protected expression, then it is arguably infringement. Hence, most AI art tends to be on the idea side of the divide unless it closely duplicates particular expressions, which would risk being in breach of copyright law.

Storage of Data by AI as Copyright Infringement

How generative AI (genAI) models retain copyrighted material while training poses a curious question: is merely retaining data, never to be exposed to users, copyright infringement? The Indian Copyright Act states, inSection 14, that the act of storing or copying a work, though in a digital form, is a privilege solely held by the owner of the copyright. What this implies is that when AI systems store books, images, or music to learn from, it could theoretically be considered copyright infringement, even if that material isn't directly displayed or used. Copyright law does have its limits, though. It protects the expression of ideas, not the ideas themselves. If AI is just analysing patterns or pulling out general information, without reproducing the creative expression, it might not be considered infringement. Courts may consider whether such storage amounts to copyright infringement or falls under the exceptions like the idea-expression dichotomy, *de minimis* use (which refers to very minimal use), or functional use (like for search or plagiarism detection) to justify such storage.

Case Law: Ankit Sahni's AI-Generated Artwork 'Suryast'

Ankit Sahni, an artist-lawyer, created 'Suryast' using AI software RAGHAV by blending his photo with Van Gogh's The Starry Night. Sahni directed inputs but the U.S. Copyright Office denied copyright registration, finding AI, not Sahni, as the creative force. Despite appeals, human contribution was deemed insufficient. In contrast, India initially granted copyright registration in 2020 but later questioned the AI tool's legal status and co-authorship validity. The registration remains on record, but procedural clarity is lacking. Canada granted protection acknowledging Sahni's input; the U.S. and India require clear human authorship. The Beijing Internet Court offers a middle path, granting protection if humans exercise control over creative elements. The case highlights the urgent need for legislative clarity on AI authorship and copyright registration transparency, especially as AI tools become common in creative processes.

Comparative Analysis of Copyright Law in Different Countries

European Union (EU)

The EU's AI Act (2024) regulates AI use by risk levels, complementing copyright law. While AI itself cannot own copyrights, human creators using AI tools retain protection. The AI Liability Directive seeks to clarify responsibility. Secondary legislation is expected to define AI authorship and copyrightability clearly.

United States (US)

The US requires human authorship for copyright. The U.S. Copyright Office rejects registrations where AI is sole creator (e.g., Thaler and Sahni cases). Human selection or editing of AI output may warrant protection. The USCO is reviewing copyright and AI guidance for registrations.

China

China recognizes AI-generated works under certain conditions, emphasizing human involvement. Enforcement and fair use exceptions for AI training remain inconsistent but evolving.

Japan

Japan permits limited copyright protection for AI-generated works with clear human involvement, treating AI as a support tool. Courts assess the degree of human input before granting protection, balancing innovation with the protection of creativity.

Conclusion

AI-generated art challenges traditional copyright frameworks on authorship, ownership, and infringement. Current laws worldwide, including India's, require meaningful human creativity, but AI disrupts these notions by producing works with minimal human input. The Suryast case exemplifies the legal uncertainty and jurisdictional divergence. Legislative clarity is urgently needed to balance innovation with protecting human creators' rights, defining liability, originality, and fair use in the AI era.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.