Why Even the Most Clever AI Inventions Fail to Pass India’s Patent Filter

why even the most clever ai inventions fail to pass india’s patent filter?

rnrn

rnrn

rnrn

even the most innovative ai inventions often fail to secure patents in india, not only because they lack inventiveness, but also due to overlooked legal and procedural pitfalls. patent applications for ai-related inventions face intense scrutiny, not just on novelty and inventive step, but also on critical requirements like sufficient disclosure, claim clarity, and patent eligibility. the case of caleb suresh motupalli vs. controller of patents (madras high court, 2025) highlights how patent applications can stumble on grounds such as insufficient technical detail, vague claims, omitted best modes, or improper amendments. this article explores why ai inventions fail at the indian patent office and provides actionable insights to avoid these common pitfalls, with reference to the caleb suresh motupalli case and key provisions of the patents act, 1970.

rnrn

rnrn

rnrn

understanding the invention that got rejected
rnrnthe invention in caleb suresh motupalli case aimed to address the loss of human agency and control due to increasing ai capabilities. titled "necktie persona-extender/environment-integrator and method for super-augmenting a persona to manifest a pan-environment super-cyborg," it proposed a system to non-invasively integrate human intelligence with ai, creating a "super-augmented persona" or "human 2.0." key features included:

rnrn

rnrn

rnrn

persona extender/augmenter: a system to extend and augment human capabilities using ai.
rnrnblack-box modernization: techniques to integrate human intelligence with ai systems.
rnrnlabourspace: a layered environment combining physical ("meatspace") and virtual ("cyberspace") elements.
rnrnchristocratic necked service-oriented architecture (cnsoa): a governance system for managing interactions between humans and ai.

rnrn

rnrn

rnrn

despite its ambitious scope, the invention was rejected due to several legal shortcomings, which serve as valuable lessons for ai patent applicants. let's examine each requirement and the specific issues that led to rejection.

rnrn

rnrn

rnrn

1) enablement and sufficiency of disclosure (section 10(4)(a))
rnrnunder section 10(4)(a) of the patents act, a complete specification must fully and particularly describe the invention and how to perform it, enabling a person skilled in the art (psita) to work the invention without undue experimentation. in caleb suresh motupalli, the court found the ai invention’s description lacked the necessary technical detail and was overly abstract, making it impossible for a skilled person to implement without undue experimentation.

rnrn

rnrn

rnrn

the patent application for a human-ai "persona-extender" was rejected in part because critical operational elements were not explained, and the promised results could not be achieved from the disclosure. for example, the invention relied on black-box modernization techniques and distributed object technology (dot), but the specification failed to provide sufficient details on how these techniques could be applied to achieve the claimed results.

rnrn

rnrn

rnrn

best practice: draft the specification with clear, complete examples and embodiments of how the ai works. provide algorithms steps, flowcharts, training data details, or pseudo-code if applicable. ensure that every feature in the claims is supported by sufficient how-to information in the description. if your ai solution involves interdisciplinary techniques, explain each aspect for a person skilled in that field. in short, disclose enough so others can actually carry out your invention.

rnrn

rnrn

rnrn

2) clarity and conciseness in claims (section 10(5))
rnrnsection 10(5) requires that the claims be clear, succinct, and fairly based on the disclosed matter. vague or overly verbose claims not only confuse the scope of protection but also risk rejection. in caleb suresh motupalli, the patent claims were extremely long, filled with abstract and undefined concepts (even biblical metaphors), making them unclear and too broad. the court agreed that the claims lacked a clear technical feature and were not properly anchored in the description, violating section 10(5).

rnrn

rnrn

rnrn

best practice: keep claims focused and unambiguous. use straightforward terminology (or define unusual terms in the specification). each claim should ideally express one inventive concept in a single sentence that isn’t unwieldy. avoid piling on excessive jargon or results-oriented language that doesn’t concretely limit the invention. ensure every claim element has support in the description – this is the fair basis aspect.

rnrn

rnrn

rnrn

3) best mode requirement (section 10(4)(b))
rnrnindian patent law insists on disclosure of the best method of performing the invention known to the applicant at the time of filing. section 10(4)(b) embodies this requirement. the idea is that inventors must not hold back their preferred way of implementing the invention while seeking monopoly rights.

rnrn

rnrn

rnrn

in caleb suresh motupalli, the applicant argued that certain claims reflected the best mode of the invention, but the court noted the complete lack of any working example or technical detail to actually achieve the claimed "persona augmentation" in practice. the absence of any workable criteria or concrete steps for the key technical features meant the best mode was effectively not disclosed at all.

rnrn

rnrn

rnrn

best practice: always include at least one working example or embodiment that represents the best way you know to implement the ai invention. this could be an example training configuration for a machine learning model, a preferred network architecture, or specific parameter values that yield optimal results. make it clear in the description that this is the preferred embodiment.

rnrn

rnrn

rnrn

4) patent eligibility of ai inventions (section 3(k))
rnrnperhaps the most notorious hurdle for ai and software-related applications in india is section 3(k), which excludes "a mathematical or business method or a computer program per se or algorithms" from patentable subject matter. the key is the phrase "per se" – indian courts have interpreted it to mean that while pure software or abstract algorithms are not patentable, an invention having a technical application or technical effect is not merely a software per se.

rnrn

rnrn

rnrn

in caleb suresh motupalli, the court applied this test and found the claimed human-ai integration system lacked any demonstrated technical effect on hardware or any tangible improvement – it remained an abstract idea and was deemed ineligible.

rnrn

rnrn

rnrn

best practice: when drafting ai-related patent applications, emphasize the technical application and outcomes. frame your invention as a solution to a technical problem. for example: does your ai algorithm improve the speed or accuracy of image recognition beyond a normal computer’s capability? highlight such benefits in the description. tie the algorithm to a specific apparatus or system, such as "a vision system for autonomous vehicles" or "a medical diagnosis device implementing [the algorithm]."

rnrn

rnrn

rnrn

5) scope of amendments and fair basing (section 59(1))
rnrnanother legal pitfall comes during prosecution: amending claims beyond what was originally disclosed. section 59(1) mandates that no amendment shall be allowed that "claims matter not in substance disclosed in the specification" as filed, and any amended claim must fall wholly within the scope of the original claims. in caleb suresh motupalli, the applicant made claim amendments, but the patent office held that some amended claims introduced concepts that were not originally disclosed, thus contravening section 59(1).

rnrn

rnrn

rnrn

best practice: draft the initial application with a forward-looking eye, including sufficient fallback positions and alternatives so that you have basis for possible amendments. during prosecution, any claim amendments should be checked against the original description – every added element or broadened scope must have literal or at least implicit support in the filed spec.

rnrn

rnrn

rnrn

closing notes
rnrnbeyond the headline tests of novelty and non-obviousness, the indian patent office and courts demand that ai-related patent applications meet rigorous standards of disclosure and form. a successful ai patent application should read as a teaching document: enabling the invention (with the best mode), delineating the claims with clarity, demonstrating a technical contribution (to escape the section 3(k) exclusion), and remaining faithful to its original disclosure through any amendments.

rnrn

rnrn

rnrn

as the caleb suresh motupalli case shows, overlooking any of these aspects can be fatal. for patent applicants in the ai arena, the takeaways are clear. invest time in the patent drafting stage to pre-empt objections: describe your ai invention in depth, exemplify the best mode, craft concise claims that align with that description, and explicitly highlight technical effects or improvements. by doing so, you greatly improve the chances that your ai patent application will circumvent its examination successfully and emerge as a granted patent.

rnrn

rnrn

rnrn

disclaimer: this article is provided for informational purposes only and should not be construed as legal advice. the analyses presented are based on our interpretation of recent case law and may not reflect all legal perspectives. readers are encouraged to consult with qualified patent attorneys for advice on specific patent matters. for personalized guidance on ai patent applications in india, please contact us at mail@radeyip.com
rnrn 

rnrn

rn

rn