A deeptech venture usually begins life with one asset that matters: a piece of knowledge that took a laboratory years to produce. Everything that follows, from the first grant to the first acquisition conversation, depends on whether that knowledge has been turned into something the company owns and can defend. A deeptech IP strategy is the plan for doing that. It decides what gets patented, what stays secret, what gets published, and in what order, so that the founders do not spend scarce money protecting the wrong things or, worse, lose rights through an early disclosure they did not think about.
Most founders treat intellectual property as a filing exercise handled by an attorney once the product is ready. In deeptech that is too late and also too narrow. The clock on novelty starts the first time an idea leaves the lab, and the choice between a patent and a trade secret shapes the business model, the investor conversation and even hiring. This article sets out how the Indian system works for an early-stage science company, where the timing traps lie, and offers a filing decision tree organised by technology type and stage.
This is a practitioner's framework, not legal advice. Every filing decision should be made with a registered patent agent or IP counsel who has read the actual invention disclosure.
Why deeptech IP strategy differs from software IP
In a typical software company, the defensible asset is speed, distribution and data. Code is protected by copyright automatically, and few early-stage software founders in India file patents at all. A deeptech company is different in three ways.
First, its core asset is often a physical or scientific mechanism: a new material, a sensor architecture, a molecule, a fabrication step, a control method for hardware. These can be reverse engineered once a product ships, which makes secrecy fragile and patents more valuable.
Second, the timeline is long. A patent filed at the laboratory stage may have used a meaningful share of its twenty-year term before the company earns its first rupee. Filing too early can waste term; filing too late can mean losing novelty to a conference talk or a thesis.
Third, investors in deeptech price IP explicitly. Because there are few traction metrics in the early years, a diligence team looks at what the company owns, whether the founders or their university own it, and whether the claims would stop a well-funded competitor. A separate PanScience Research piece covers research commercialisation and university licensing; this article assumes the company either owns its IP or has a clean licence, and asks what to do next.
How the Indian patent system treats an early-stage venture
The Indian Patent Office has become noticeably busier and somewhat faster. WIPO reported that patent filings in India rose 19.1% in 2024, the sixth consecutive year of double-digit growth. The office of the Controller General of Patents, Designs and Trade Marks (CGPDTM) recorded 110,375 filings in 2024-25, of which 61.79% were domestic, according to its annual report as summarised by S.S. Rana. The patent office's own achievements note states that the time to issue a first examination report fell to 10.6 months in 2023.
For startups, three concessions matter.
The first is cost. The patent office states that an 80% fee concession is available to startups, MSMEs and educational institutions for patent filing, prosecution and maintenance, with 50% off trademark fees and 75% off design fees. For a company recognised by DPIIT, the official fee for a full Indian prosecution is therefore a small fraction of what a large entity pays. Attorney fees, drafting and foreign filing remain the real cost.
The second is speed. Startups are eligible to request expedited examination. In its 2019-20 summary, the patent office reported that under the expedited system most applications were decided within about one year of the request, with the fastest grant taking 41 days. That matters for a venture that needs a granted patent before a strategic negotiation or a funding round.
The third is facilitation. The Start-ups Intellectual Property Protection (SIPP) scheme, launched in 2016, pays empanelled facilitators so that startups receive drafting and filing support without paying professional fees directly. The patent office reports that in April 2023 the scheme was extended to 31 March 2026. Founders should confirm with the patent office whether it has been renewed beyond that date before relying on it.
The timing rules that decide most outcomes
Most expensive IP mistakes in deeptech are timing mistakes. The rules below are drawn from the Patents Act and Rules as described by Obhan and Associates and S&A Law Offices, and from WIPO's guidance on the Patent Cooperation Treaty (PCT).
Two of these deserve emphasis.
The Patents (Amendment) Rules 2024, notified on 15 March 2024, cut the deadline for requesting examination from 48 months to 31 months. A founder who planned to keep an application pending for four years while the product matured no longer has that option. Examination now happens on a timeline closer to that of Europe.
Section 39 is the trap that catches researchers who collaborate internationally. An Indian resident may not file a patent application abroad for an invention without written permission from the Controller, unless an Indian application for the same invention was filed at least six weeks earlier and no secrecy direction applies. Permission is sought on Form 25 and is ordinarily decided within 21 days. S&A Law Offices notes that breach can lead to the Indian application being treated as abandoned and to imprisonment of up to two years, a fine, or both. For a founding team split between an Indian lab and a foreign co-inventor, the order of filings has to be planned before anyone presses submit.
The PCT itself is often misunderstood. WIPO is clear that there is no international patent: the PCT lets an applicant file one application with effect in its 159 contracting states and receive an international search and written opinion, but grant remains a national decision. Its real value for a young company is time. It defers the large national-phase costs to month 30, by which point the company should know which markets matter.
Patents, trade secrets and the Indian gap
The core strategic choice is between disclosure in exchange for a time-limited monopoly, and secrecy for as long as it can be kept. India adds a complication: there is no trade secrets statute. The 22nd Law Commission, in its 289th report, recommended dedicated legislation and proposed a Protection of Trade Secrets Bill, 2024. As OA Obhan Mason noted in May 2024, the draft is recommendatory, and until Parliament acts, trade secrets in India are protected through contract, breach of confidence and equitable principles. In practice that means a deeptech venture relying on secrecy must do the work a statute would otherwise do: written confidentiality and assignment terms with every employee, intern and contractor, access controls on lab notebooks and code, and a register of what is treated as secret.
The choice depends on whether the invention can be seen in the product.
Software and AI inventions face a further filter. Section 3(k) of the Patents Act excludes mathematical methods, business methods, algorithms and computer programmes per se. The draft revised Guidelines for Computer Related Inventions published by CGPDTM in March 2025, as analysed by SS Rana and Co in Bar and Bench, set out how examiners should read that exclusion. Following the Ferid Allani and Microsoft line of cases, subject matter implemented on a general-purpose computer is not rejected as a programme per se if it produces a technical effect, such as higher speed, reduced memory access, better control of a robotic arm, improved signal transmission or real-time control of devices. The same commentary notes the Madras High Court's view that disclosure of an AI invention must be specific enough for the model's functionality to be reproduced. For an AI venture, the implication is that a claim drafted around the technical problem in a physical system has a far better chance than one drafted around the model.
PanScience's portfolio offers a useful illustration of the split. The studio describes Proteus as an LLM gateway that routes, secures and monitors model traffic, and PredCo as industrial AI and predictive maintenance for manufacturing plants. A gateway's value sits mostly in software behaviour that competitors cannot see, a natural fit for secrecy and speed. Industrial AI that changes how a machine is monitored or controlled touches the technical-effect test directly, which is where a patent claim can be built. These are categories of decision, not a description of either venture's filings.
A filing decision tree by technology type and stage
The framework below is the original contribution of this article. It treats IP as a sequence of decisions tied to technology readiness and commercial stage, rather than a single filing event. A separate PanScience Research piece maps technology readiness levels to commercial milestones; the stages here use the same logic in simpler terms.
Step 1. Before any external disclosure, ask four questions.
1. Is there a concrete, reproducible result, not just a hypothesis? If no, keep lab records and confidentiality, and do not file yet. 2. Is a disclosure scheduled (paper, thesis, conference, grant report, pitch deck sent widely)? If yes, file a provisional application in India before it, regardless of stage. 3. Will the invention be visible in a product or detectable in a competitor's? If no, default to trade secret and document it. 4. Is any inventor resident outside India, or is a foreign filing planned first? If yes, plan the Section 39 sequence now.
Step 2. Choose the default instrument by technology type.
Step 3. Match spending to stage.
Step 4. Review annually. At each anniversary, prune patents that no longer map to the product, check the trade secret register against who has left the company, and re-run the four questions for new work.
The logic underneath the tree is simple. File early only to protect priority against a known disclosure. File broadly only when a market is real. Keep secret what cannot be seen. Spend on drafting before spending on countries.
What investors and partners look for
Diligence on a deeptech company usually asks a short list of questions. Who owns the IP: the company, the founders personally, or the institution where the work was done? Are there signed assignments from every inventor, including students and former employees? Do the claims cover the product the company is actually selling, or an earlier version? Has anyone searched for third-party patents in the first target market? Is there a written trade secret policy?
The Indian rules make some of these easier to answer than they were. The new certificate of inventorship under Rule 70A lets an inventor obtain a certificate naming them, for a fee of ₹900, which can ease tension with academic co-inventors when the company holds the patent. The shift of the working statement to a three-yearly cycle reduces compliance noise. But none of this replaces the basic hygiene of assignments and records.
For corporate partners, the issue is improvements. A pilot agreement that says the customer owns anything developed during the engagement can strip a young company of its most valuable follow-on inventions. A separate PanScience Research piece covers how corporate IP licences to new ventures should be structured, including field-of-use and improvement clauses.
Common mistakes
Five patterns recur in early deeptech companies.
Filing a provisional application that is too thin. A provisional only protects what it discloses. A two-page note filed the night before a conference may not support the claims needed a year later.
Treating the PCT as protection in itself. It is a deferral mechanism. The money arrives at month 30, and companies that have not planned for it either abandon applications or spend money they need for product.
Ignoring Section 39. Teams with foreign co-inventors or a foreign university partner can create a filing that cannot be cured.
Relying on secrecy without contracts. In the absence of a statute, a trade secret that is not backed by written obligations is very hard to defend.
Patenting the algorithm instead of the effect. For AI and software inventions, claims built around a technical result in a physical or computing system are far more likely to survive Section 3(k).
Conclusion
A deeptech IP strategy is a set of sequenced choices, not a stack of filings. The Indian system now offers startups real advantages: an 80% fee concession, expedited examination and facilitation support. It has also tightened in ways founders need to notice, especially the 31-month deadline for examination requests and the long-standing Section 39 permission rule. The absence of a trade secrets statute means that secrecy in India is only as strong as a company's contracts and records.
The decision tree above reduces the problem to four questions before disclosure, a default instrument by technology type, and spending matched to stage. Used with a competent patent agent, it should stop most young companies from making the two errors that cost the most: losing novelty before filing, and spending on foreign protection before the market is known.
Methodology and sources
This article combines statutory and procedural rules from official and professional sources with an original filing decision tree. The decision tree, the patent versus trade secret table and the stage-based spending table are illustrative frameworks built by PanScience Research; they are not drawn from a dataset and are not legal advice. Fee concessions, expedited examination outcomes, SIPP history and first examination report timing come from the Indian Patent Office's achievements note (ipindia.gov.in/writereaddata/Portal/Images/pdf/ipoAchievements.pdf). Filing growth is from WIPO press release PR/2025/943 and the CGPDTM Annual Report 2024-25 as summarised by S.S. Rana (editor- verified). Prosecution timelines are from Obhan and Associates, "Timelines in Prosecuting an Indian Patent", Mondaq, 27 May 2022 (mondaq.com/india/patent/1196542/timelines-in-prosecuting-an-indian-patent), updated by Obhan and Associates, "Patent (Amendment) Rules, 2024: Key Changes", Mondaq, 26 March 2024 (mondaq.com/india/patent/1442682/patent-amendment-rules-2024-key-changes). Section 39 requirements are from S&A Law Offices, Mondaq, 13 May 2021 (mondaq.com/india/patent/1066932/section-39-of-the-patents- act-1970). PCT facts are from WIPO's PCT FAQs (wipo.int/pct/en/faqs/faqs.html). The trade secrets position is from OA Obhan Mason, "Law Commission Proposes a Draft Bill on Trade Secrets for India", Mondaq, 17 May 2024 (mondaq.com/india/trade-secrets/1465858/law- commission-proposes-a-draft-bill-on-trade-secrets-for-india). The treatment of computer related inventions is from SS Rana and Co, "Revised CRI Guidelines 2025", Bar and Bench, 7 April 2025 (barandbench.com/law-firms/view-point/revised-cri-guidelines-2025), which discusses the March 2025 draft. PanScience venture descriptions are company-reported from panscience.xyz and are used only to illustrate categories of decision.