Generative AI & Issues of Copyright

Generative AI and Copyright: A Collision That Copyright Law Can’t Ignore

The advent of Generative AI exposes a glaring inadequacy in copyright law—an inability to adapt to a technological disruption that blurs the lines between fair use and unauthorized exploitation. AI’s reliance on copyrighted material for training datasets doesn’t just trespass legal grey areas; it challenges the foundation of intellectual property rights altogether.

The Institutional Landscape: Copyright Law’s Historical Limitations

Copyright law in India, governed by the Copyright Act, 1957, was crafted in an era devoid of non-human creators. Section 2(d) of the Act rigidly defines “author” as a human entity, effectively excluding AI-generated works from its purview. While the government has clarified its stance against creating separate IPRs for AI-generated content, this legislative rigidity fails to address the complexities of “fair use” or database rights in the age of AI.

Globally, the frameworks differ substantially. The Berne Convention, to which India is a signatory, ensures mutual recognition of copyright protections but offers no technical prescriptions for machine-generated works. In contrast, Japan has explicitly exempted AI training data from infringement claims, provided that the usage is non-consumptive—aimed solely at machine learning. This proactive interpretation offers a model India could emulate, but the domestic legal landscape remains trapped in interpretive ambiguity.

Undisclosed Datasets: Evidence of Exploitation

Central to the controversy is the method of dataset creation. Generative AI models like ChatGPT and MidJourney are trained on datasets scraped indiscriminately from the internet, including copyrighted content. Although OpenAI’s introduction of an opt-out mechanism appears conciliatory, it merely bypasses accountability for past infringements—a fact emphasized by lawsuits against OpenAI in the Delhi High Court. Notably, Bollywood music labels and news agencies, such as Federation of Indian Publishers and ANI, allege theft of intellectual property without consent.

The economic ramifications are vast. For instance, copyrighted media forms a significant chunk of India’s creative economy, valued at over ₹15,000 crore annually. Generative AI’s unchecked absorption of such works not only undermines creators but also threatens jobs dependent on copyright royalties. This becomes evident in the accusations leveraged by Indian journalists and artists who see AI platforms cannibalizing their intellectual property for profit.

Countering ‘Fair Use’: A Tale of Regulatory Capture?

The strongest argument in favor of Generative AI rests on the doctrine of “fair use,” as championed by AI corporations in U.S. courts. OpenAI, for example, asserts that training its models constitutes “fair learning in education,” aligning generative AI with the public good. However, this defense overlooks the for-profit nature of such models. The claim of operating “for educational purposes” fails when the end-product is sold as a subscription-based service, benefiting private profits rather than public knowledge.

More critically, this narrative feeds into a pattern of regulatory capture, where technology giants lobby for overly lenient regulations under the guise of innovation. In India, the Delhi High Court has rightly questioned whether AI’s ability to “forget” copyrighted material—a concept akin to unlearning—is technically feasible. This goes directly to the heart of enforcement challenges, pushing for technological solutions alongside legislative oversight.

International Comparisons: Lessons from Japan

While India grapples with ambiguity, Japan’s copyright provisions take a pragmatic stance. Under Japanese law, AI training using copyrighted data is exempt from infringement claims if deemed non-consumptive. This approach balances innovation and intellectual property, providing legal clarity while fostering technological growth. What Japan frames as “non-consumptive learning,” India currently interprets through piecemeal litigation, creating significant uncertainty for both developers and creators.

What India calls “enumerated exceptions” for educational use barely scratches the nuances required for AI regulation. Japan’s distinction enables algorithmic development without jeopardizing human authors’ creative incentives—an equilibrium India’s courts and policymakers have yet to achieve.

Challenges to My Position: Can Copyright Law Keep Up?

To argue that traditional copyright law alone can regulate Generative AI would understate its limitations. Global legal ambiguity demonstrates that no jurisdiction has perfected AI ownership regulation. Critics might suggest that stricter copyright enforcement risks stifling innovation in a vital technology frontier.

Undoubtedly, imposing stringent penalties or access restrictions could disincentivize AI firms from operating within Indian borders, a blow to India's burgeoning AI-driven economy. However, the counterweight to these concerns lies in transparency measures, such as public opt-out registries and fair licensing agreements. These mechanisms protect creators while fostering technological progress, thereby providing a middle ground that preserves both innovation and intellectual property.

Assessment: A Legal Reset Is Overdue

India’s copyright framework is standing at a legislative crossroads. Institutional remedies must go beyond punitive litigation to include proactive policy reforms such as AI transparency laws, binding opt-out mechanisms, and fair remuneration for dataset usage. These aren’t just beneficial; they’re essential to realign the copyright ecosystem with the realities of algorithmic creativity.

What is clear is that copyright law must evolve to serve both its original mandate—protecting creators—and the emerging mandate of equitable AI regulation. Borrowing lessons from Japan, India has the opportunity to lead an ethical AI governance model that respects human ingenuity while unlocking machine potential.

Practice Questions for UPSC

Prelims Practice Questions

Frequently Asked Questions