Copyright protects the expression of a human mind. So what happens when the expression comes from a system whose output its own programmer cannot predict, let alone explain? India answers by asking who caused the work to be created — which turns out to be a harder question than it sounds.
What artificial intelligence is
Put simply, AI is non-biological intelligence. But what is intelligence? There is no conclusive definition, and over the years no agreement even among AI researchers. Painting broadly, intelligence means the ability to solve complex problems.
In the AI context, intelligence divides into narrow and general:
- Artificial Narrow Intelligence (ANI) — the ability to solve a single or narrow range of tasks extremely well, often better than humans. IBM’s Deep Blue defeated chess master Garry Kasparov in 1997, the first machine to surpass humans at a specific task. Interestingly, Deep Blue could not have beaten a teenager at tic-tac-toe.
- Artificial General Intelligence (AGI) — machine intelligence equivalent to human intelligence, spanning the full panoply of human skills. As at the time of writing, AGI did not exist, and researchers were sceptical about its arrival in the coming decades, given the complexities involved — not least that we build toward human-like intelligence with an incomplete understanding of how human cognition works.
An important aspect is machine learning — a branch of AI focused on building applications that learn from data and improve their accuracy over time without being explicitly programmed to do so. Familiar examples: Siri, streaming recommendations, chatbots, self-driving cars.
Where copyright comes in
Copyright subsists in a work the moment it is expressed in a tangible form, and protects the expression of an idea, not the idea itself. The work must involve the creativity and originality of a human being for the law to consider it for protection.
The US Copyright Office has stated the position succinctly: it will register a work of authorship only if the work is created by a human being.
That position sits awkwardly with the technology. ANI is already capable of producing music, literary pieces and video games on its own through machine learning — output that can rival human work in form, if not in provenance.
The Indian and UK approach
India and the United Kingdom take a different route: they grant copyright to the person who causes the work to be created.
The UK vests rights over computer-generated works in the arranger, under Section 9(3) of the Copyright, Designs and Patents Act:
“In the case of a literary, dramatic, musical or artistic work which is computer-generated, the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken.”
India, under Section 2(d) of the Copyright Act, 1957, grants copyright in a computer-generated work to the person who causes the work to be created — which could be the programmer who wrote the software, or the owner of the software.
Why that answer strains
Complexities arise in determining the author. ANI creates work based on the code — models and algorithms — written by the programmer, and on the datasets fed to it. And ANI can and does generate output that the programmer cannot explain.
That produces a series of questions the law has not answered:
- How can copyright be granted to the programmer as author when they are uncertain of the output the system will generate?
- What happens if the generated work is against the law, contrary to the programmer’s intention?
- What if the software is not owned by the programmer and generates unlawful content due to faulty programming?
There are no answers, because there are no rules in India governing the functionality of ANI.
Before rules can be laid down, it is essential to understand the technical realities of examining ANI from a computer science perspective. It is difficult to draw legal boundaries around a technology without first understanding its abilities and functions.
What has happened since
This was written before generative AI became ubiquitous, and the questions it raised have been tested since — mostly confirming that the hard part was correctly identified.
- The US position has held. In Thaler v. Perlmutter, the courts upheld the Copyright Office’s refusal to register a work listing an AI as author, affirming the human authorship requirement. The Office’s subsequent guidance allows registration of works containing AI-generated material, but only for the portions reflecting human authorship, which must be disclosed.
- India encountered it directly. The Copyright Office registered a work listing an AI application as a co-author alongside a human, then issued a withdrawal notice — leaving the position unresolved rather than settled.
- The UK reconsidered Section 9(3). The provision that once looked forward-thinking has been the subject of consultation on whether to reform or repeal it, on the view that it protects output without a human author in a way that sits oddly with copyright’s rationale.
- The training question overtook the authorship question. The most consequential litigation now concerns whether using copyrighted works to train models infringes — a question this piece did not reach, and which has become the central commercial issue.
The core difficulty identified here remains unresolved: attributing authorship to a person who cannot explain or predict what the system produced. Every jurisdiction has either required a human author and excluded the output, or deemed an author by statute and left the deeming increasingly strained.
The takeaways
- Copyright protects human expression — the US registers only works created by a human being.
- India and the UK deem an author — the person who causes the work to be created.
- Unexplainable output breaks the deeming — you cannot meaningfully author what you cannot predict.
- India still has no rules governing AI functionality — the gap identified here persists.
Frequently asked questions
Can an AI be the author of a copyrighted work? Not in the United States, which requires human authorship. India and the UK instead assign authorship to the person who causes the work to be created.
Who owns copyright in a computer-generated work in India? Under Section 2(d) of the Copyright Act, 1957, the person who causes the work to be created — potentially the programmer or the owner of the software.
What does Section 9(3) of the UK CDPA say? That for computer-generated literary, dramatic, musical or artistic works, the author is the person by whom the arrangements necessary for creating the work are undertaken.
Is AI-assisted work registrable? In the US, only the portions reflecting human authorship, which must be disclosed. India’s position remains unsettled after a registration listing an AI co-author drew a withdrawal notice.
Useful official resources
- The Copyright Act, 1957
- Copyright Office, India
- WIPO — Artificial Intelligence and IP