India’s AI & IP Strategy: Driving Innovation and National Advantage

Contents

Executive Summary

This panel discussion addresses the critical intersection of Artificial Intelligence policy and Intellectual Property rights in India's development as a global AI leader. The conversation balances the urgent need for AI innovation with the complex challenges of protecting copyright, data rights, and creator livelihoods—ultimately concluding that while IP protections matter, innovation should be the overwhelming priority for India at this stage of its AI journey.

Key Takeaways

India Faces a Binary Choice: Either create broad copyright exemptions for AI training (democratizing innovation) or establish mandatory licensing with government compensation (protecting creators). There is no middle ground—the government must choose which rights to protect and which to subordinate.
Three Data Domains Need Different Rules: Public GenAI training data, proprietary enterprise models, and individual privacy data cannot be governed by a single framework. India should develop differentiated policies for each layer rather than a one-size-fits-all law.
Technology Enforcement Matters as Much as Legal Rights: Watermarking, provenance tracking, and automated infringement detection are as critical as laws themselves. India and the US should collaborate on these tools rather than focusing only on legal definitions of ownership.
Institutional Support Accelerates Startup Confidence: Startups need pre-vetted datasets, democratized compute, lab-to-market support, and talent trained in both AI and policy/law—not just legal clarity. Removing barriers is as important as defining rights.
"Less Is More" for IP Regulation: Rather than comprehensive IP frameworks that may slow innovation, India should surgically address documented harms (deepfakes, impersonation, privacy violations) while letting the market and courts resolve edge cases. Perfect IP law can be the enemy of necessary innovation.

Conference Talk Summary

Key Topics Covered

AI and IPR Framework: How to reconcile monopoly protections (IP) with competition law; role of AI in IPR administration
Generative AI and Copyright: Copyright infringement concerns, creator protections, and the "vampirical" nature of training on existing content
Data Ownership & AI Training: Who owns training data, AI-generated outputs, and the implications for startups and businesses
Regulatory Approaches: Comparison between legal frameworks vs. technology-driven governance; India's potential as a regulatory leader
Enterprise vs. Public Data: Distinction between proprietary enterprise data, public commons data, and individual privacy
Open Source & Innovation: Balancing open-source benefits with IP protections for downstream creators
Enforcement & Detection: Watermarking, provenance tracking, and automated infringement detection as emerging IP enforcement tools
Text & Data Mining Exception: Legislative exemptions in copyright law to enable AI model training
Startup Support & Legal Clarity: Institutional mechanisms to help startups navigate unclear IP landscapes

Key Points & Insights

AI's Five Dimensions in IPR Context: Dr. S. Chakravarthi framed AI tools for IPR administration around data, technology, innovation, and trust—emphasizing AI should augment rather than replace human judgment in patent searches and copyright enforcement.
The "Ownership-Value-Link" Problem: Rajini highlighted that without clarity on IP ownership, AI leaders cannot accurately calculate ROI or demonstrate value—creating a recursive risk where organizations cannot confidently invest in AI infrastructure.
Three Distinct AI Layers Require Different IP Approaches: Nathan Benaich argued that GenAI (macro/universal level), enterprise proprietary data, and individual privacy layers each need distinct governance:
- GenAI Layer: Nearly impossible to control; becoming a collective asset
- Enterprise Layer: Proprietary knowledge is the differentiator; quality over quantity matters
- Individual Layer: Acute need for international data privacy standards
Mandatory Licensing vs. Text & Data Mining Exemptions: A fundamental policy choice exists between NASCOM's proposal for copyright exemptions (like Japan, Singapore, EU) allowing unfettered AI training, versus DPI's proposal for mandatory blanket licenses with government-set royalties and centralized funds.
The "Vampirical" Copyright Problem: Generative AI systems scrape pre-existing creative content without consent, damaging artists' livelihoods. Creative sectors view this as content theft rather than innovation—demanding opt-out mechanisms and impersonation rights.
Enforcement Gaps Between Law and Technology: Sabine Malik (Rapid 7) identified a critical gap: legal frameworks alone are insufficient without technological enforcement tools (watermarking, provenance tracking, automated detection). Cyber security and IP enforcement are converging.
Quality Data as Competitive Advantage: In an AI-saturated world where raw data is abundant, the real differentiator becomes high-quality, proprietary datasets—described as "a pinch of salt in a bowl of soup" that gives taste to generic models.
Opt-Out Mechanisms & Impersonation Rights: Jennifer Mulveni (Adobe) proposed two concrete protections: (a) creatives' ability to opt-out of training on their work, and (b) a new impersonation/personality right allowing legal recourse when AI systems exploit someone's likeness for economic gain.
The Government's "Leap of Faith" Decision: Shardul articulated that governments must choose whether to negate pre-existing copyright rights at scale to enable universal AI development—creating exceptions to Section 52 of copyright law, effectively telling rights-holders "your rights no longer apply."
Innovation Over Perfection: Nathan Benaich's closing argument: India should be "surgical" about IP enforcement, addressing extreme cases (deepfakes, impersonation, enterprise privacy, consumer data protection) while prioritizing innovation and growth over comprehensive IP regulation that may stifle development.

Notable Quotes or Statements

"IPR creates monopoly rights and competition law frowns on monopoly. How to marry them?" — Dr. S. Chakravarthi (on the fundamental tension between IP and antitrust)

"Patient interests need to trump patent interests. Patients are more important than patents." — Dr. S. Chakravarthi (on balancing public welfare with IP rights)

"Trips agreement should not be allowed to trip public interest." — Dr. S. Chakravarthi (on international agreements and domestic priorities)

"The whole link problem I call it the ownership-value-link problem." — Rajini (on why unclear IP ownership prevents ROI calculation in AI projects)

"In a world moving from data-scarce to data-abundant, how can you say somebody owns it? It's like universal consciousness." — Nathan Benaich (on why traditional IP concepts may not apply to GenAI)

"Innovation is the overwhelming imperative. Less is more." — Nathan Benaich (closing argument on regulatory prioritization)

"IP is digital, global, and globally distributed. Legal recognition alone just isn't enough. Enforcement requires innovation in the legal regime." — Sabine Malik (on the enforcement-technology gap)

"There's a right for impersonation and a right to say 'do not train on my work'—these are starting points, not silver bullets." — Jennifer Mulveni, Adobe (on practical creative protections)

"The government is effectively asked to negate the rights which have been created over time and make it an open field." — Shardul (on the policy dilemma governments face)

Speakers & Organizations Mentioned

Panelists

Dr. S. Chakravarthi: Former civil servant (40-year career, retired 30 years prior), author on IPR and competition law, keynote speaker
Rajini: AI leader at a GCC (Global Capability Center); expertise in AI ROI and enterprise ML
Sabine Malik: Rapid 7 (cybersecurity firm); focus on IP enforcement and digital infrastructure
Nathan Benaich: innoGo (enterprise AI company); strategic advisor on AI policy and enterprise data
Jennifer Mulveni: Adobe; government relations and public policy in APEC region; focus on creative industries
Shardul: IP and copyright law expert; detailed analysis of legislative options
IBM Representative (IBM Granite models mentioned): Open-source and foundation model focus

Organizations & Entities

NASCOM: Host organization; positioned on copyright exemptions for AI training
DPI (Digital Public Infrastructure): Proposed mandatory blanket licensing model
IBM: Open-source AI models (Granite family); mentioned as transparent foundation model provider
Adobe: Creative software, content credentials, provenance tracking tools
Rapid 7: Cybersecurity and IP enforcement
innoGo: Enterprise AI and proprietary data governance
Delhi High Court: Case of ANI vs. OpenAI (currently pending on authorship/ownership)
Government of India: Electronics and IT Ministry (Minister Ashini Vishna mentioned); developing India's AI mission

Referenced Frameworks & Standards

NIST AI Framework: Governance reference mentioned
Singapore's AI Verify: Governance reference mentioned
TRIPS Agreement: World Trade Organization's Trade-Related Aspects of IP Rights
India's IT Act, IT Rules, Data Protection/Privacy Acts: Existing legal foundations

Technical Concepts & Resources

AI Concepts

Generative AI (GenAI): Large language models and text-to-image systems creating new content from training data
Large Language Models (LLMs): Trained on scraped internet data; core of copyright concerns
Synthetic Data: AI-generated training data; raises quality and recursion risks if used to train new models
Foundation Models: Base models (e.g., IBM's Granite family) designed for transparency and auditability
Text & Data Mining (TDM): Process of automatically extracting patterns from large datasets; central to copyright debate

IP & Legal Concepts

Copyright/Fair Dealing vs. Fair Use: Distinction between jurisdictions; core of training exemption debate
Section 52 of Copyright Act (India): Defines permitted uses; proposed exception area for TDM
Patent Search & Prior Art Detection: AI can accelerate by scanning global patent databases at lower cost
Personality Rights / Impersonation Rights: Emerging legal category addressing deepfakes and likeness theft
Content Credentials & Provenance Tracking: Adobe tool enabling creators to mark and track usage restrictions
Watermarking: Embedding ownership/consent information in digital content
Automated Infringement Detection: Tools to identify unauthorized AI training or output usage

Data Governance Concepts

Big Data's Five V's: Velocity, Volume, Value, Variety, Trustworthiness (Dr. Chakravarthi's framework)
Proprietary Enterprise Data: High-quality, differentiated data creating competitive advantage
Public Commons Data: Universal training datasets treated as collective assets
Individual/Personal Data: Privacy-sensitive data requiring explicit consent frameworks
Mandatory Blanket Licensing: DPI proposal: central fund with government-set royalty rates, non-negotiable
Opt-Out Mechanisms: Creative's ability to prevent training on their work (Adobe's content credentials)

Policy Frameworks Mentioned

India's AI Mission: National strategy emphasizing safety, deepfake detection, bias detection, model unlearning
Japan, Singapore, EU Frameworks: Referenced as having TDM exemptions or opt-out copyright provisions
NIST & Singapore's AI Verify: Governance models for transparency, explainability, auditability

Organizations & Case Law

ANI vs. OpenAI: Delhi High Court case on authorship, ownership, and liability (still pending; unresolved)
India's Role as Regulatory Leader: Mentioned as potential model between US legal approach and technology governance approach

Contextual Notes

Conference Setting: NASCOM-convened AI summit in India; emphasis on US-India partnership and India's positioning as a global AI innovation leader.

Regulatory Landscape: India is actively formulating AI and copyright policy; this discussion reflects live policy deliberation rather than settled law. The government's final decisions on copyright exemptions vs. mandatory licensing remain pending.

Urgency: Multiple speakers emphasize that startups and enterprises need clarity now to invest confidently, even as perfect legal frameworks take years to develop.