[What ethical and legal safeguards should govern how AI companies source training data? Suggest a regulatory framework applicable to India. (GS-II, 15 marks)](/upsc-mains-answer/ethical-legal-safeguards-should-govern-ai-9341e8e)

Reports that AI firms have been destructively scanning purchased physical books — cutting bindings, digitising pages and discarding the remains — to feed large language models have made training-data sourcing a live governance question [1]. Since data is the raw material of AI, its acquisition must be disciplined by both ethics and enforceable law.

Ethical safeguards required

• Consent and provenance: creators and data subjects should know that their work or personal data enters a training corpus; sourcing from pirated repositories is ethically indefensible [1].
• Transparency: dataset composition and acquisition methods should be disclosable, not revealed only through litigation [1].
• Cultural stewardship: digitisation must not destroy rare or irreplaceable works — preservation is a public trust, echoing NITI Aayog's "Responsible AI for All" principles of safety and public good [4].
• Fair compensation: authors and publishers whose works build commercial models deserve a share of value.

Legal safeguards required

• Copyright clarity: India's fair dealing exception under Section 52, Copyright Act 1957 was drafted for research and review, not industrial-scale machine ingestion; text-and-data-mining needs an express, bounded exception [3].
• Data protection: the Digital Personal Data Protection Act, 2023 requires lawful, consent-based processing — its application to scraped personal data must be clarified through rules [2].
• Accountability: statutory duty to maintain auditable dataset records, with penalties for pirated-source training.
• Judicial testing: the pending ANI–OpenAI litigation before the Delhi High Court will shape Indian doctrine.

A framework for India Build on the India AI Governance Guidelines, 2025 (MeitY, IndiaAI Mission), which favour a light-touch, sectoral, accountability-based approach [5]: a mandatory dataset disclosure register, a licensing-cum-royalty mechanism for copyrighted works, techno-legal audits, and grievance redress under existing regulators rather than a new omnibus law.

India can convert this global controversy into first-mover advantage — a framework that rewards lawful, documented data sourcing will protect creators' livelihoods and cultural heritage while sustaining the innovation the IndiaAI Mission seeks, aligning technology with constitutional guarantees of privacy and dignity.