Context

• Copyright law has historically evolved in response to technological innovations.
• From its inception in 1710, born out of the printing press revolution, to modern-day digital complexities, copyright has aimed to strike a balance between protecting creators and promoting public access to knowledge.
• Today, the emergence of generative artificial intelligence (AI) has reignited a familiar debate: Can copyright law keep pace with technological advancements, or does it risk becoming obsolete in the face of rapidly evolving AI capabilities?

Historical Context and Evolving Challenges

• Origins of Copyright: A Response to the Printing Press
  • The evolution of copyright law is inseparable from the history of technological innovation.
  • Its origins can be traced to the Statute of Anne in 1710, widely regarded as the first modern copyright law.
  • This legislation was a direct response to the invention of the printing press, a revolutionary technology that democratized access to information but also introduced the risk of unauthorised reproduction.
  • The law was designed to grant limited rights to authors and publishers to control the use of their works, thereby encouraging learning and rewarding creativity while preventing exploitation.
• The Digital Disruption: Internet and Global Access
  • The digital age marked a profound shift. With the rise of the Internet, the scale and speed of content distribution exploded.
  • Suddenly, texts, images, videos, and music could be copied and shared globally with a single click.
  • This necessitated more robust copyright regimes and the development of digital rights management (DRM) tools and anti-piracy legislation.
  • Yet, even in this era, the fundamental legal focus remained on the reproduction and distribution of works, the tangible copying of protected material.
• A Paradigm Shift: Learning, Not Copying, with Generative AI
  • Now, with the arrival of generative artificial intelligence, we are witnessing a new paradigm.
  • Technologies like Large Language Models (LLMs) and image generators do not function by simply copying or redistributing content.
  • Instead, they learn from massive datasets, many of which include copyrighted works, to create entirely new outputs.
  • This form of learning, often described as training, represents a departure from traditional conceptions of copying.
  • The AI does not necessarily store or replicate specific original works; it analyses patterns, structures, and relationships within data to generate novel content.
• Copyright’s Resilience to Technological Change
  • Importantly, such technological and legal disruptions are not unprecedented.
  • Similar anxieties have surfaced with almost every major technological breakthrough.
  • For instance, when cassette tapes became popular, the music industry feared rampant piracy.
  • When VCRs entered the market, film studios worried about unauthorized recording of broadcasts.
  • Yet, in each case, copyright law found ways to evolve, either through legislative amendment, judicial interpretation, or the creation of new licensing models.
• The Unseen Complexity of AI: Legal and Ethical Implications
  • In the case of AI, however, the complexity is deeper. The learning process is invisible, algorithmic, and distributed across vast datasets.
  • This opacity makes it harder to trace infringement or apply existing legal tests.
  • Moreover, the sheer scale of content ingestion, often conducted without the creators’ knowledge or consent, poses ethical and economic dilemmas about ownership, credit, and compensation.

Generative AI, Legal Crossroads, Jurisdictional Complexities and Legal Variations

• Generative AI at Legal Crossroads
  • Companies like OpenAI have come under scrutiny for their training practices, which involve Internet scraping, an automated process where Large Language Models (LLMs) are trained on both copyrighted and non-copyrighted material.
  • This method has sparked global legal challenges. In India, the Federation of Indian Publishers and Asian News International have filed infringement suits against OpenAI, alleging unauthorised use of their works.
  • Similar legal actions are unfolding in the United States.
  • Notably, OpenAI has responded by introducing an “opt-out” mechanism, enabling content owners to exclude their materials from future training sets.
  • However, this policy does not address past training, which remains a contentious issue.
  • In the Indian context, Professor Arul George Scaria, serving as amicus curiae, has emphasised the need for courts to consider the feasibility of unlearning copyrighted content and to ensure that AI development does not come at the cost of access to legitimate information.
• Jurisdictional Complexities and Legal Variations
  • A significant hurdle in resolving these disputes lies in the varying interpretations of copyright exceptions across jurisdictions.
  • Unlike the U.S., which employs a broad ‘fair use’ doctrine that includes provisions for educational use, India’s Copyright Act takes an enumerated approach.
  • Exceptions are explicitly listed and narrowly defined, limiting flexibility. In India, education-related exemptions are confined to classroom use, making it harder for AI companies to claim legitimate training rights.
  • This stricter legal framework may work in favour of rightsholders in India, but it also risks obstructing access to knowledge, ironically counteracting the original purpose of copyright.
  • Furthermore, the opt-out mechanism proposed by OpenAI might create a divide between well-established AI platforms with extensive datasets and emerging players who may lack access to high-quality training materials.
  • Courts must, therefore, consider the need for a level playing field in the generative AI landscape.

The Way Forward: Reframing Copyright for the Future

• At the core of this debate is a broader philosophical question: Should copyright law differentiate between human and machine learning?
• Human creativity has always built upon existing works, each generation learning from the last.
• Generative AI, in many ways, mimics this process. However, existing legal frameworks do not distinguish between the outputs of human and machine creators, leading to tensions in interpretation and enforcement.
• A more sustainable solution lies in returning to the foundational principles of copyright. The law protects the expression of ideas, not the ideas or facts themselves.
• As long as AI systems are using existing information to learn, without replicating the original expression, they are not necessarily infringing on copyright.
• When AI outputs begin to mirror or closely mimic protected works, the current legal mechanisms are equipped to respond appropriately.

Conclusion

• Generative AI poses novel questions for copyright law, but it does not render it obsolete. Rather, it invites a re-examination of the law’s scope and purpose.
• A balanced approach, grounded in the core tenets of copyright, sensitive to jurisdictional differences, and attuned to the realities of AI training, can help bridge the gap between innovation and regulation.
• Courts and policymakers must ensure that copyright continues to protect creators without stifling the evolution of technology or creativity itself.