AI & Copyright Law

Core Issue

1. At the heart of the debate: Are generative AI models built on copyrighted works?
2. Two major US copyright lawsuits (Writers vs. Anthropic & Authors vs. Meta) question whether AI training violates copyright law.

Source : The Indian Express

Relevance : GS 2(Governance ) ,Gs 3(IPR , Technology)

Court Verdicts So Far

1. Writers vs. Anthropic (Aug 2023, US)

1. Authors including Michael Chabon, George R.R. Martin filed case.
2. Accused Anthropic of copying copyrighted texts for training Claude AI.
3. Court’s response: Did not rule on copyright infringement directly; stated the AI model does not “transform” texts enough to qualify as fair use.

2. Authors vs. Meta

1. Authors sued Meta for training LLaMA models using their copyrighted books.
2. Judge dismissed part of the complaint on procedural grounds.
3. However: Meta could still be held liable if models “memorise and regurgitate” copyrighted content.

Key Legal Concepts

1. Fair Use: Permits limited use of copyrighted work without permission if it transforms the content (e.g. parody, research).
2. Transformative Use: AI must add new expression or meaning to qualify.

The Case Against OpenAI in India

1. In 2024, ANI and the Indian Music Industry (IMI) accused OpenAI of:
   1. Training models on copyrighted Indian content.
   2. Violating Section 65A of Indian Copyright Act (on circumvention of tech protections).
2. No judgment yet. Jurisdiction under question.

Challenges in India

1. OpenAI operates in India, but:
   1. No clarity on how training data is sourced.
   2. AI firms claim only “public” data is used.
   3. Lack of explicit Indian law on AI & copyright.

Why It Matters

1. India’s creative industry (books, music, cinema) is at risk of unauthorised AI replication.
2. Worries about AI models “memorising and regurgitating” original work.
3. Raises ethical & legal questions around ownership and consent.

Global Implications

1. Courts have not yet conclusively ruled if using copyrighted work for AI training is legal.
2. Verdicts will set precedents for AI governance globally, affecting OpenAI, Meta, Google, Anthropic, etc.

Significance of Rulings

1. So far, courts have favoured tech companies but have not shut the door on future liability.
2. If plaintiffs prove “verbatim memorisation” by models, it could trigger compensation or licensing models.

Key UPSC-Relevant Facts from the Article

Key Dimensions:

Legal Gaps in India

• No AI-specific copyright law.
  • Copyright Act, 1957 doesn’t define AI authorship or fair use for training data.

Ethical Concerns

• Use of creative content without consent or credit.
  • Undermines originality and creator rights.

Economic Impact

• Threat to livelihoods of artists, authors, musicians.
  • Monetization of pirated or public content by AI companies.

Technology vs Regulation

• Balance between fostering innovation and protecting IP.
  • Ambiguity over “transformative use” of copyrighted material.

Global Comparisons

• US: “Fair use” doctrine allows AI training.
  • EU: Tight opt-outs under TDM rules.
  • Japan: Broad AI training exemptions.

Regulatory Vacuum

• No guidelines under IT Act or DPDP Act for AI training data.
  • Digital India Act still pending.

Privacy and Consent

• Training data may include personal content without consent.
  • Conflicts with data protection principles.

Creator Rights & Royalties

• Lack of collective bargaining tools (e.g., CMOs for AI usage).
  • No attribution mechanism for original creators.

AI Liability & Accountability

• Who is responsible for AI-generated infringements — developer or deployer?
  • No legal clarity yet.

Public Good vs Private Profit

• Use of public domain data for private AI profit.
  • Debate over open-source mandates for public-trained models.