On August 21, 2026, the Department for Promotion of Industry and Trade (DPIIT) published a Draft Manual on Trade Marks: Practice & Procedure 2026 (Draft Manual) in the Gazette of India, inviting public comments within fifteen days. Running over two hundred pages, the Draft Manual is meant to be the operational rulebook for examiners and registrars and a guide for prospective applicants of a mark. The Draft Manual is an updated version of the 2015 Manual. It has added a new chapter on well-known marks, laying down the process for registration and, for the first time, attempting to address issues around the consent of a person whose representations have been generated by AI for the purposes of a mark.
The attempt to modernise the process is welcome, but a close reading reveals inconsistencies and, at times, a lack of explanation for certain procedures, with some practices being legally imprecise. The following heads elaborate on the same, beginning with the timeline given for public comment.
A Comment Window Too Short to Matter
What makes the fifteen-day window particularly hard to defend is what the Draft Manual actually does. This was not a routine update. The Draft Manual was revised after eleven years. It accounts for technological developments, formalises digital filing and introduces an entirely new chapter with the Draft Manual running over two hundred pages in total. Yet it was compressed into the same fifteen-day window typically reserved for far narrower amendments. In that time, a reader is expected to cross-check the Draft Manual against the Trade Mark Manual 2015, Trade Marks Act 1999 and the Trade Marks Rules 2017, verify the case laws cited, assess the sensitive areas it has touched upon such as AI-generated representations and a formalised process for well-known marks and prepare a reasoned submission, which is hardly enough time.
When guidelines governing something as consequential and technically dense as pharmaceutical and biotech patent examination receive a 15-day window, shorter than the routine amendment rules in 2023 and 2024 it’s difficult to see how meaningful, well-researched feedback from practitioners, civil society and industry can realistically be compiled and submitted in time. The effect, whether or not that’s the intent, is a consultation process that looks more like a formality than genuine engagement. A sixty-day minimum window for guidelines of this scale isn’t an unreasonable ask. It’s closer to the baseline that other jurisdictions treat as standard practice for stakeholder input on technical rule-making.
A Term Without a Definition: “AI-generated Representations of a Person”
The most conceptually significant addition in the Draft Manual is paragraph 2.6.1, which offers an office action against anyone using an “AI-generated representation or any identifiable attribute of a natural person in the trademark” (page 43) to attach a written and signed document of consent from that person or their legal representative. The underlying intent, to keep the Registry practice current with how generative AI can be used to appropriate a person’s persona, is sound and timely. But whether introducing ‘AI-generated representations’ within the Manual was appropriate can be debated. This paragraph/clause is concerned with representation of a person being used without his other consent. Whether the representation is hand painted or artificially generated or designed, doesn’t matter. Categorising representation to be AI generated and non AI-generated was unnecessary.
And inclusion of ‘AI generated representation’ gives rise to further questions about how this would play out.
First, it is unclear whether “representation” as used in Section 14 and Rule 29 of the existing framework was ever intended to cover AI-generated or AI-assisted output. Both provisions predate the generative AI era, and the Draft Manual does not explain the interpretive basis for its extension of “representation” to synthetic content. Even if it did intend to categorise representations based on mediums, the only intention behind the rule is to ensure consent of the person whose representation is made in the logo was taken prior to filing an application for the trademark.
Second, “AI-generated representation” is neither defined in the Act, the Rules, nor the Draft Manual itself. Does it cover only depictions wholly generated by AI or also AI-assisted edits of an authentic photograph, composite or ‘inspired-by’ imagery, or AI-generated voice and likeness more broadly? Without a working definition, examiners have no consistent basis for assessing such representations and applicants have no way of assessing compliance in advance.
The two judicial authorities the Draft Manual relies on do not resolve this uncertainty as well, if anything, they compound it. Anil Kapoor v. Simply Life India is related to AI-generated representations as it expressly addresses AI-generated deepfakes, ringtones and GIFs. DM Entertainment v. Baby Gift House by contrast, predates the generative AI context by a decade. Citing it as an authority for specifically AI-focused requirements extends a general publicity rights doctrine into a technological context that the decision never contemplated. This is a larger interpretive leap than the Draft Manual’s citation suggests and requires further clarification.
Two further practical questions merit clarification. Does the Registry have the technical capacity to detect AI-generated content in a submitted mark during examination, and if not, should applicants instead be required to declare the use of AI when filing? Separately from consent, where an applicant uses a third-party tool to generate a mark, does the applicant actually hold proprietary rights over the output, given that the AI tools’ terms of use may restrict or condition such rights? This is a gap separate from the consent issue, but it is worth highlighting because it points to the practical difficulties stakeholders are likely to face under the revised version of 2.6.1.
Separately, CGPDTM recently issued Draft Guidelines on Use of AI in Proceedings Before the TM Registry (“Draft Guidelines”) on the directions of Delhi High Court in the case of Stryten Energy LLC v. Nidhi Agarwal & Ors (interested readers can refer to Kaustub’s recent post on this here). The Draft Guidelines primarily deal with use of AI in preparing documents and citing case laws. It is unclear whether the Draft Guidelines are intended to reach AI-generated representation of a natural person forming part of a mark. It would be welcoming if the Office could clarify this. The Draft Guidelines however are not confined to pleadings and citations. Clause 2.2 applies to every application, “image” and “label” filed before the Registry at the stage of filing and clause 3.1(b) extends AI-generated content to synthetically generated audio-visual material. The representation of the mark, which is annexed to the application, is itself an image filed before the Registry. It should therefore fall within clause 2.2. If the logo looks AI-generated, clause 14.4 lets the Registrar ask for the original file and its creation details since such a logo could harm the rights of the person whose likeness is shown. The Draft Manual already treats AI-generated representation of a natural person in a mark as an examination concern under 2.6.1. The Draft Manual and the Draft Guidelines should work together with a disclosure requirement on the application where a mark contains an AI-generated or AI-altered likeness and a clear route for the Registrar to seek proof of consent.
Can a Prior User’s Right Always Be Considered Superior?
Chapter 7 of the Draft Manual (pages 195-215) deals with well-known marks and states that a “prior-user’s rights can always be considered superior” (page 195). This is perhaps an overstatement which does not accurately reflect the law. There is no blanket rule of absolute preference for a prior user, envisaged under Section 34 nor Rule 25(2). Section 34 does not grant prior users an unqualified right of superiority, it rather provides that a registered proprietor cannot interfere with a person who has been using an identical mark. Rule 25(2) merely requires an applicant claiming prior use to file an affidavit along with supporting documents to provide a proof of the same. Hence, the chapter’s language is at odds with the statute it is meant to build upon and risks misguiding both the Registry examiners and applicants, including those opposing a well-known mark. The practical implications are worth flagging here.
First is that the claim also glosses over two qualifications that were inserted to accommodate other possibilities. Section 12 expressly allows a bona fide, honest and concurrent user to register an identical or similar mark which the chapter does not acknowledge. Secondly, it ignores the possibility of acquiescence provided under Section 33, which provides a limitation to the rights of prior users if they fail to oppose a later registered user’s use of the mark for a continuous period of five years. Omission of these possibilities while asserting an absolute claim of preference to a prior user creates a false sense of certainty and gives a person no reason to even ask whether the period of acquiescence has passed. It would be appreciated if the Draft Manual could be cross-referenced with Sections 12, 33, and 34 for the same.
Why Must Objections to Well-known Marks Go Only to Mumbai Office?
One final concern is a procedural inconsistency with apparently no stated justification. Rule 4 of the Trade Marks Rules 2017, does not require filing at any particular office. The appropriate office is ordinarily determined by the principal place of business of the registered proprietor, which the Draft Manual itself confirms in paragraph 1.1 (page 7), also noting that the documents can be filed electronically through the IP India website.
Yet the newly added Chapter 7, on well-known marks, requires opposition statements to be filed specifically at the Mumbai Office with no explanation for the departure. Rule 4 draws no distinction between types of applications and makes no special provision for well-known marks. A Manual which is meant to operationalise the Rules cannot carve out an exception that the Rules do not contain, and that too without any explanation. It becomes even more inconsistent when the rest of the Draft Manual moves towards digital acceptance.
Larger Point
None of this suggests that the Draft Manual’s underlying goals are misguided. Updating the Registry practice to account for AI-generated content and consolidating scattered documents are genuinely appreciable practices. But a Manual that examiners and applicants will rely on for years will be bound by, needs precision, internal consistency with the Act and Rules and a comment process long enough to catch such gaps before they are enforced.
Disclaimer: The views expressed here are the author’s own!
