The Human Behind the Machine: What India's DABUS Order Means for AI and IP Ownership

India’s DABUS ruling confirms IP law protects human creators, not AI tools. Learn how human direction secures copyright ownership and manages legal risk. 

Who owns what an AI creates? For years intellectual property law has answered a version of this question by looking past the tool — the camera, the ghostwriter, the research assistant — to the person who directed it. Artificial intelligence is simply the newest, most capable entrant on that list. A recent order from the Indian Copyright Office, refusing to register an AI system called DABUS as the "author" of an artwork, confirms that this old answer still works. IP law does not need reinventing for the AI age — it needs its human-centred foundations reaffirmed, and the DABUS order does exactly that.

AI Is a Tool, Not a Creator

International IP law has always rewarded human judgment. The Berne Convention's moral rights and the TRIPS Agreement's "inventive step" requirement both assume a human mind capable of choice and creative leaps. AI, however sophisticated, performs computation — it has no "intent" and no "judgment" of its own. What it does have is speed: it can generate thousands of variations in the time a human takes to draft one. But someone still has to define the problem, evaluate what the AI produces, and decide what is worth publishing, filing or selling. That someone is where the law should — and largely does — locate ownership.

Copyright: The Human's Guiding Hand Still Counts

Under most copyright regimes, a human's prompts, curation and editing can supply all the originality the law requires. Courts already reward this kind of contribution: in Eastern Book Co. v. D.B. Modak (2008), the Supreme Court of India protected work built on skilful selection and arrangement, not raw invention. In 2023, the Beijing Internet Court went further in Li v. Liu, holding that an AI-generated image was copyrightable because the user's "intellectual investment" in configuring the model and picking the final image was enough. The US took a stricter line: in Thaler v. Perlmutter, the D.C. Circuit held that only humans can be "authors" under the US Copyright Act, and refused registration because the applicant had listed an AI system, not himself, as author. The Supreme Court declined to hear a further appeal in March 2026, leaving that ruling in place. Notably, the courts found that Dr. Thaler had never properly argued that he himself was the author — a mistake, as it turns out, he was about to repeat in India.

The DABUS Order: India Draws a Careful Line

The work at the centre of the Indian order is the very same artwork litigated in the US: "A Recent Entrance to Paradise," generated by Dr. Stephen Thaler's AI system, DABUS. Thaler applied to register the work in India too, again naming DABUS — not himself — as author. On 31 August 2026, the Copyright Office rejected the application, but its reasoning is worth close attention.

First, the Registrar held that the work is original enough to be copyrighted. Even though DABUS's neural network generated the final image, Dr. Thaler had curated the photographs and text fed into the system and set the generative process in motion — enough to clear copyright law's modest originality bar.

Second, and more significantly, the Registrar interpreted Section 2(d)(vi) of the Copyright Act, 1957 — India's specific rule for "computer-generated" works, which names the author as "the person who causes the work to be created." Borrowing a "mastermind" test from older US case law on photography and film, the Registrar found that Dr. Thaler — who conceived DABUS, fed it the inputs, and triggered the process — was, on his own account, the person who caused the work to be created. DABUS merely executed the final step.

So why was the application still rejected? Because Dr. Thaler, even when given the chance during the hearings to amend his application and name himself as author, refused. He insisted DABUS remain listed as author. The Registrar made clear that this — not any blanket rule against AI-assisted authorship — is why the application failed. Indeed, the order leaves the door open: had Dr. Thaler named himself as author, the Registrar's own findings suggest the application would likely have succeeded.

Who's Liable When AI Gets It Wrong?

If ownership tracks human control, so should liability. Placing responsibility on the person who directs and publishes an AI's output — rather than the AI itself — removes any temptation to use AI as a shield against infringement claims. The same principle applies to misuse of a person's image or voice, such as deepfakes: Indian courts have already stepped in to protect individuals against unauthorised commercial use of their likeness, as in Anil Kapoor v. Simply Life India (2023), using ordinary passing-off and publicity-rights principles rather than any AI-specific law. Just as a knife manufacturer isn't liable when its knife is used to commit a crime, an AI developer shouldn't be liable for how a user chooses to deploy its tool.

Patents: The Same Rule Applies

Patent law tells the same story. Inventorship requires a "mental act of conception," something current AI cannot perform on its own. DABUS's patent applications failed for the same reason in the US and the UK: courts in both jurisdictions held that only a human can be named as an inventor. AI can accelerate research and suggest solutions, but it is the human who recognises which solution is actually useful and turns it into a legally valid claim. That human should sign the application — and carry the legal risk if the claim later fails.

Trademarks and Contracts: Who Bears the Risk

AI-generated logos raise similar issues, but the underlying rule doesn't change: trademark rights come from use and registration, not from who — or what — designed the mark. The business that adopts and uses an AI-generated logo is responsible for clearing it, not the AI tool that produced it. The same logic should apply commercially: AI developers ought to remain responsible for problems in their underlying models, such as training-data infringement, while businesses using AI tools should bear responsibility for how they use the output. Well-drafted contracts and indemnities are the right place to allocate that risk — not a search for someone, or something, else to blame.

What This Means for Your Business

• If you use AI tools to create content, keep a record of your prompts, inputs and edits — that record is what may establish you, not the tool, as the legal author.

• When filing a copyright or patent application for AI-assisted work, name yourself as the author or inventor, and be ready to explain your role in directing the process.

• Build human review into any workflow that uses AI for creative or inventive output, and document it.

• Make sure your contracts with AI vendors, employees and clients clearly allocate responsibility for AI-generated content and any resulting IP disputes.

The Bottom Line

The DABUS order is a reminder, not a revolution. Legislators may one day decide that AI deserves its own category of rights — the Registrar was careful to say that this remains Parliament's call, not the Copyright Office's. Until then, the safest and most legally sound position for any business is the one the law has taken all along: AI is the tool, and the human who directs it is the owner — and is accountable for what it produces.