Download PDF
Download PDF
GS Linked
PYQ Based

AI Training and the Right to Privacy

GS Paper 2: Government Policies; Fundamental Rights | GS Paper 3: IT and Computers; Awareness in AI

Large AI models are trained on vast amounts of personal data, much of it scraped from the open internet. India''s data protection framework was not designed for this use, leaving gaps that weaken a fundamental right.

Core Argument

The DPDP Act, 2023 is built on consent and purpose limitation. AI training breaks both: data is gathered without consent, and models can memorise or infer personal details long after training ends.

The law must therefore be extended explicitly to AI training.

  • The Legal Framework
  • In K.S. Puttaswamy v. Union of India (2017), the Supreme Court held privacy to be a fundamental right under Article 21. Any restriction must meet the tests of legality, necessity and proportionality.
  • The DPDP Act, 2023 and the DPDP Rules notified in November 2025 set up consent-based processing, duties for Data Fiduciaries and a Data Protection Board.
  • The older IT Act, 2000 and its 2011 SPDI Rules penalise negligent handling of sensitive data with up to three years in prison or a Rs 5 lakh fine.

Gaps in the Law

  • The DPDP Act has no separate category for sensitive data such as health, biometric or financial information.
  • Data that a person has made publicly available falls outside the Act, so scraping for AI training is largely unregulated.
  • Models can reproduce or deduce details such as health, sexuality or location that appear in no single input. This inference risk is not addressed.
  • Key obligations are phased in only by May 2027. Unlike the EU''s GDPR, India gives no right to object to AI training. A broad research exemption can also be stretched to cover commercial model training.

Way Forward

  • AI training, including on scraped public data, should be brought explicitly within the DPDP framework. A sensitive personal data category with stricter consent rules is needed.
  • Users should get a right to object and opt out. The research exemption should be narrowed, and developers should disclose their training datasets.
  • Oversight should align with the IndiaAI Mission''s safety institute and a risk-based approach to AI governance.

Conclusion

Innovation and privacy need not be traded off. A clear rulebook for AI training will build public trust, itself a precondition for India''s AI ambitions.

Mains Practice Question: "India''s data protection law was written for data processing, not for machine learning." Examine the privacy risks of AI model training and suggest legal reforms. (GS Paper 2, 15 marks, 250 words)

Posted on 04-11-2026 • By Admin

Coming Soon

Student Self Tracker app will be available soon.