Artificial Intelligence and Machine Learning: Innovation and IP Protection at IIT Madras

Artificial Intelligence (AI) and Machine Learning (ML) continue to redefine the way we live, work, and innovate. On February 19, 2025, DexPatent organized an online Video Conference for the faculty, research scholars and students of IIT Madras to explore the evolving landscape of AI/ML technologies, with a special focus on innovation and intellectual property protection. The conference was well attended and engaging.

1. Latest Trends in AI/ML

AI is moving beyond narrow, task-specific models toward more general-purpose systems. Leading the charge are multi-modal AI, generative models, and self-supervised learning. Machine learning models are increasingly embedded into edge devices, enabling real-time decisions without depending on cloud infrastructure. Another key shift is the growing emphasis on responsible AI, focusing on explainability, bias mitigation, and regulatory compliance.

2. New Advancements and Key Players

Tech giants like OpenAI, Google DeepMind, Meta, Anthropic, and NVIDIA continue to push boundaries. Breakthroughs such as GPT-5, Gemini Ultra, and Sora are unlocking new possibilities in natural language processing, video generation, and general intelligence. Meanwhile, startups and academia are driving innovation in specialized fields like healthcare AI, climate modeling, and neuro-symbolic AI.

3. Patenting Trends in AI/ML

Globally, there’s been an exponential rise in AI-related patent filings, particularly in sectors like healthcare diagnostics, autonomous systems, and AI hardware optimization. India too, is witnessing increased patent activity, as both academia and startups realize the value of protecting AI-driven innovations.

4. What Can Be Patented in AI/ML?

Inventions with technical applicability and tangible outcomes are typically patentable. Examples include:

• Novel neural network architectures for specific applications
• AI models improving industrial processes
• Embedded AI in hardware systems
• AI-driven control mechanisms with real-world utility

Algorithms as such may not be patentable, but their application in a technical process or product certainly can be.

5. Common FER Objections and How to Overcome Them

First Examination Reports (FERs) often raise the following objections like:

• Lack of Novelty (Section 2(1)(j))
• Lack of inventive step
• Non-patentability under Section 3(k) (mathematical or business methods)
• Section 3(m) – Non-patentability of Methods of Performing Mental Acts or Playing Games
• Insufficient disclosure
• Objection of Broad claims

To overcome the above-mentioned Objections, a few of the following need to be taken care:

Clearly show technical effect and industrial application

Include detailed descriptions, flow diagrams, and real-world implementation examples

Emphasize how the AI/ML solution solves a technical problem, rather than just automating a business rule

This interactive session not only showcased cutting-edge AI developments but also emphasized the critical role of protecting such innovations through robust patent strategies. As IIT Madras continues to be at the forefront of technological research, understanding both the science and the IP framework will be key to shaping the future of AI in India and beyond.