When the Designer Is a Machine: Authorship, Novelty, and Ownership in AI-Generated Industrial Designs

Introduction : A decade back, turning a rough sketch into a set of designs that looked real usually took weeks. One person would draw and then the group would debate the options. After that, a maker or a renderer would turn the chosen idea into a form a client could actually review and  now the clock has changed and the Generative tools can cut that wait fast as With one text prompt, an AI aided CAD add on, a parametric system, or an image generator may spit out many possible shapes. You can get chair looks, phone case drafts, packaging previews, and even app screen mockups. All of it can happen before anyone finishes a cup of coffee. Companies that work in furniture, consumer tech, and packaging have already adjusted their process and their path from brief to prototype now runs at this new speed. As a result, Indian design law, written for a time when a person did the drawing, is starting to feel the pressure. “”

The real issue is who gets credit, not whether the result is good and the output can look neat and exact, almost like it was made with care by a human hand yet still, it was not drawn in the usual way. In practice, someone enters the prompt, ignores most of the responses, and then edits and refines the single image they like. Even then, it is unclear who should be treated as the designer. It is also unclear whether that person can seek registration under the Designs Act, 2000. The law does not spell out the answer in a clear way and this is not a puzzle that only comes up in India. Other countries have dealt with similar disputes. 

Those rulings were not about factory style or industrial form and they also did not come from an Indian court. Still, both point straight at the issue an Indian applicant has to face today. This matters for business, not only for study or theory and it affects whether a product shape that starts from a prompt can be registered in the first place. It also affects whether that registration can be used later if someone else copies the same shape. On top of that, it affects whether the design can be sold or licensed when the question of who owns it is still not settled and this text looks at four linked questions that grow out of that same blank space. First, who counts as the applicant when a person gives the prompt and a system then produces the result. Next, what generative tools do to both novelty checks and the risk of later disclosure. Then, how AI platforms split responsibility for ownership of what is produced. Finally, why does a registered design give a tighter kind of protection than copyright or patent law.”

Legal Provisions

A. The Designs Act, 2000 (India)

The law starts in Section 2(d). It says a “design” is about visible features such as shape, layout, pattern, or ornament and these features have to be put on an article through an industrial method. Then Section 2(g) speaks about “originality.” And It links originality to the person who made or created the design. The text also covers a familiar design when it is used for the first time on a new type of article.  Next, Section 2(j) defines the “proprietor.” It is the person who commissioned the design and paid for it and It can also be someone who later got it by assignment. Taken together, these parts point to a human role behind the design. Someone either makes it, or hires another person to make it. The wording does not deal with a situation where a design is produced without a human maker.  

Section 4 lists reasons registration may be refused and One reason is that the design is not new or not original and Another reason is that the design was revealed anywhere in the world before the filing date. Registration can also be blocked if the design is too close to something already known, or if the design mixes known elements without enough difference. The law also rejects designs that are scandalous or obscene. Section 5 requires the application to be filed by “any person claiming to be the proprietor.” That wording again assumes a real person makes the claim.  Section 21 gives a small exception to the disclosure rule. If a design is shown at a recognised exhibition, that showing will not ruin novelty. But the Controller must be told in a proper way, and the application must be filed within six months of the exhibition.

B. Relationship with the Copyright Act, 1957

The Copyright Act shows a sharp difference and Under Section 15(2), a registrable design loses copyright once an item that uses it has been made in industry more than fifty times and At that point, the Designs Act is supposed to handle the matter, not copyright. There is also another piece in the Copyright Act. Section 2(d)(vi) covers computer made works. It sets authorship in favor of the person who put in place the steps needed to make the work and By contrast, the Designs Act does not have a matching rule for designs made by machines. Because of that, the same authorship problem that copyright law answers by statute is still open in design law. In practice, it depends on how courts interpret Sections 2(j) and 5.

C. International Framework: The WIPO Hague System

People outside India can use the Hague Agreement to file one international application and they file it through the World Intellectual Property Organization and then they can seek protection in several contracting states, including India. Under the Common Regulations for this system, the applicant must be either a natural person or a legal person. That rule was set long ago, before generative AI became a topic in everyday practice. It does not really address a design whose exact appearance was made by a machine. The machine used a set of instructions from a human.

D. The Patent Statutes Behind the DABUS Litigation

The Designs Act does not say much about work made by a machine and  because of that gap, people look to patent rules for a nearby match and the issue about who counts as an inventor has already been argued in three places, and those cases turn on very similar statutory wording and In the UK, section 7 of the Patents Act 1977 limits the right to apply to the person who actually devised the invention. It also covers that person's successor in title. UK courts have treated that phrasing as pointing to a real person. In the US, the Patent Act uses a clear label. Under 35 U.S.C. § 100(f), an inventor is the “individual” who invented the subject matter. In Australia, the Patents Act 1990 uses language that also centers on a person when naming an inventor on a full application. These three laws, and there is nothing like them in the Designs Act 2000, do not spell out what to do when the deriving is done by a machine, not by a person. That is why the rulings in DABUS, which interpret those provisions, are treated in the next section as helpful but not required. They are used to guide how section 2(j) and section 5 of the Designs Act should be read.

E. Registration Practice: Locarno Classification and Examination

“A key extra step comes up in filings that use AI tools. In India, each design application is put into the Locarno Classification. This is the system run by WIPO. It sorts articles into classes and smaller sub-classes. The goal is to support the Controller’s prior-art search. In real work, the novelty check stays within the same Locarno grouping. Examiners look for earlier registrations and known designs in that class. That is also where similar material is most likely to appear. Think about product catalogues and packaging files. If AI was trained on that kind of content, the match may show up within the relevant class. So an applicant who starts with AI-created ideas should handle Locarno classification as part of the same due-care task as the novelty search. Do not treat it as a separate check. If the search is limited to the wrong class, the result is weak. A design compared to the wrong set of prior art cannot give a reliable sense of whether it can be registered.

Legal Analysis

A. Who Qualifies as the “Author” or “Applicant”?

“The Designs Act does not have a clause like Section 2(d)(vi) in the Copyright Act. So it is still unclear who counts as the “proprietor” under Section 2(j) and it could be the person who wrote and edited the prompt. It could also be the person who built the generative tool. Or it could be the party that ordered the work. No Indian court has ruled on this point for designs and still, patent cases in the DABUS line help as a comparison and  the reason is similar and  it asks whether an AI system can take a legal place that the law sets aside for a person. In Thaler v Comptroller-General, the UK Court of Appeal said the Patents Act 1977 treats an inventor as a natural person. In December 2023, the UK Supreme Court backed that view. In the US, the Federal Circuit got the same outcome in Thaler v Vidal under the Patent Act. In Australia, the Full Federal Court in Commissioner of Patents v Thaler reversed the first decision. That earlier step had said an AI matter could be treated in a way that allows naming without legal personality.

B. Novelty Searches and Disclosure Risk

“

India uses a strict rule for novelty across the whole world and Under Section 4, a design cannot have been published or used anywhere before the filing date.  One issue is the training data and these systems learn from huge collections of past work, like furniture lists, packaging records, and old product photos. So the result can look like something that already exists, even if the user had no plan to copy it. If an examiner later checks earlier material using the Locarno Classification and finds a match, that earlier item can be used to block the claim and the second issue comes from how courts describe novelty. In Bharat Glass Tube Ltd. v. Gopal Glass Works Ltd., the Supreme Court said that originality may still exist when a known design is applied to a new article. But later decisions in the High Courts have taken a narrower view. They do not treat a small surface change as enough. They also require that any earlier publication be shown in a clear form that depicts the design as it was used on a comparable article.

C. Ownership Clauses in AI Platform Terms of Service

Not every generative AI service grants the same rights. Some let the paid user take full ownership of what gets made and others keep certain “background” rights for the company and some also restrict how you may use the output unless you move to a higher plan. In many cases, free tiers come with the tightest limits. A few services also post your final work in a public gallery. That act may count as a public disclosure. It can hurt novelty if you plan to file a design claim later. Before a studio or any team files for a design, they should read the terms closely.The group also needs to confirm that it can truly claim rights in the output and They should check whether the provider holds any leftover license. If a freelancer or an outside designer is involved, the contract must be clear. The deal should state what happens to AI assisted work. This reduces the risk of missing links in the chain of rights. It also supports what is required under Section 2(j) without leaning on guesswork.

D. The Possibility of Refusal

A rejection under Section 4 is not some far off worry and  With AI helped forms, it can happen in a way that it is easy to predict. These tools are often built from public data that overlaps. So two people can use different systems and still end up with forms that look very close. Then, once one version is shown, the other version can be treated like prior material and used against it. Another issue is volume and If you generate a lot with little review, you may get text or images that should not be there. Image systems in particular can be uneven when it comes to keeping out unwanted material. That makes it more likely an output will cross into the “scandalous or obscene” area and The best practical response is proof. Keep a clear record made at the time. Save the prompts you used. Note other options you did not pick. List the exact changes you made before you used the final draft. If an examiner, or a rival, says the work is too similar, this record gives you something specific to point to. It also supports the claim that a human stayed in control, not the tool alone.

E. Why Design Protection Differs from Copyright and Patent Protection

Copyright applies to an original work the moment it is written down or otherwise made stable. You do not have to file anything to get that protection and  In India, the usual span is the author’s life plus sixty years. But Section 15(2) can reduce that time. It can end protection once a registrable design has been copied for industry more than fifty times.That is one reason makers of common consumer goods often go for design registration. They do not rely on copyright for the look of the mass-produced item and Design protection falls in the middle. It covers the ornamental or visual parts of an article. The test is based only on what a person can see. The law starts with ten years. That period can be extended by five more years. It does not ask whether the design has technical value.

Relevant Case Laws

Thaler v Comptroller-General of Patents, Designs and Trade Marks [2021] EWCA Civ 1374: The Court of Appeal held that the Patents Act 1977 presupposes a natural person as inventor and a conclusion the UK Supreme Court affirmed on further appeal in December 2023 and the reasoning applies by analogy to the person-based language of Section 5 of the Designs Act, 2000.

Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022): The Federal Circuit held that the US Patent Act's reference to an “individual” inventor means a natural person and reinforcing the cross-jurisdictional refusal to treat an AI system as an author or inventor in its own right.

Commissioner of Patents v Thaler [2022] FCAFC 62: The Full Federal Court of Australia unanimously overturned a first-instance ruling and held that only an entity with legal personality can be named as an inventor, aligning Australia's position with the UK and the US.

Eastern Book Company & Ors. v. D.B. Modak & Anr., (2008) 1 SCC 1: The Supreme Court held that originality requires a “modicum of creativity” that is  beyond purely mechanical effort, a standard that applies by analogy when assessing how much genuine human contribution sits behind an AI-assisted design.

Bharat Glass Tube Ltd. v. Gopal Glass Works Ltd., (2008) 10 SCC 657: The Supreme Court held that originality under Section 2(g) of the Designs Act that can subsist in applying a known design to a new article, while confirming that genuine novelty still faces close scrutiny against any prior publication shown in tangible form.

MICROFIBRES INC OTHERS VERSUS GIRDHAR & CO & OTHERS LNIND 2009 DEL 695:The Delhi High Court held that once a work capable of registration as an industrial design is put into industrial production and it is governed by the Designs Act's registration-based regime rather than by copyright, regardless of whether design registration was ever actually sought.

GOPAL GLASS WORKS LTD VERSUS ASSISTANT CONTROLLER OF PATENTS AND DESIGNS LNIND 2005 CAL 524: The Calcutta High Court held that a document merely downloaded from a foreign patent office's website did not amount to prior publication capable of destroying novelty and prior publication must be shown in tangible form and must actually depict the design as applied to a comparable article.

Practical Implications

If you use generative tools and keep a fresh log of what people did for each design and then Write down the prompts you tried, the options you did not pick, and the exact changes you made to the final image and This helps with an ownership argument under Section 2(j). It also gives you a clear response if someone claims the result looks too similar. Before you start any campaign that may matter for a filing, review the rules for each AI tool you plan to use and pay close attention to settings like public galleries or community feeds, since some platforms show results right away. In India, the novelty test is strict and it is treated as an absolute rule. So do not wait to file once the design is settled. If you hire freelancers or outside studios and then put the AI part in the contract.

Make it clear that the work is AI-assisted, and spell out how the rights move to the party who commissioned the job and do not leave that chain to guesswork. A stacked protection plan is often the safest route. Seek design registration for the visual look and the ornamental feel of the product. File for a patent only when there is a real functional improvement. Use copyright for flat art, or for branding parts, if you are not trying to use copyright to cover the industrial design itself. Also note the Section 15(2) limit once the design is reproduced at an industrial scale. 

Conclusion

Generative AI may speed up product design ideas from weeks to minutes and Still, the Designs Act, 2000 was drafted without this kind of tool in view andThe Copyright Act gives some hint for computer-made work. It points to Section 2(d)(vi). The Designs Act does not. It asks for an application by “any person claiming to be the proprietor.” When you read that with the DABUS patent cases, the setup looks like it expects a specific claimant. It seems to mean a human, or a legal body, not a machine. That claimant should be able to show real control over how the design ends up, not just accept the first result a system produces. At the same time, India uses a strict novelty test that is applied worldwide but the issues tied to training data and to how generative systems often share outputs are not just theoretical and can affect whether an application passes the review step. Until the law is changed, or the courts address this point head on, applicants may do better by writing down the human choices behind an AI-assisted design. That is safer than assuming the issue will not be raised.”

Author: Aryan Gandhi in case of any queries please contact/write back to us via email to content@khuranaandkhurana.com or at  Khurana & Khurana, Advocates and IP Attorney

Endnotes

1. Designs Act, 2000, s. 2(d), s. 2(g), s. 2(j) (India).

2. Designs Act, 2000, ss. 4, 5, 21 (India).

3. Copyright Act, 1957, ss. 2(d)(vi), 15(2) (India).

4. Hague Agreement Concerning the International Registration of Industrial Designs, Common Regulations (WIPO).

5. Locarno Agreement Establishing an International Classification for Industrial Designs.

6. Thaler v Comptroller-General of Patents, Designs and Trade Marks [2021] EWCA Civ 1374 (United Kingdom).

7. R (on the application of Thaler) v Comptroller-General of Patents, Designs and Trade Marks [2023] UKSC 49 (United Kingdom).

8. Thaler v Vidal, 43 F.4th 1207 (Fed. Cir. 2022) (United States).

9. Commissioner of Patents v Thaler [2022] FCAFC 62 (Australia).

10. Eastern Book Company & Ors. v. D.B. Modak & Anr., (2008) 1 SCC 1 (India).

11. Bharat Glass Tube Ltd. v. Gopal Glass Works Ltd., (2008) 10 SCC 657 (India).

12. Gopal Glass Works Ltd. v. Assistant Controller of Patents & Designs, 2006 (3) CHN 188 (Cal.).

13. World Intellectual Property Organization, Hague System for the International Registration of Industrial Designs — Guide for Users.

14. Office of the Controller General of Patents, Designs and Trade Marks, Manual of Designs Practice and Procedure (India).