From Statutory Mandate to Machine Unlearning: Navigating the Right of Erasure under DPDP
As India's DPDP Act moves toward enforcement, organizations face a new challenge: ensuring that personal data can be removed not only from databases but also from AI models that may have learned from it. Machine Unlearning is emerging as a critical capability at the intersection of AI, privacy, governance, and regulation. This roundtable will bring together experts to examine the technical, legal, and operational implications of implementing the right to erasure in AI systems. The discussion will address the realities of compliance, verification, model performance trade-offs, and the future of privacy-preserving AI in an increasingly regulated environment.
Key discussion points:- Interpreting the "right to erasure" for AI systems under the DPDP framework
- Can models truly forget? Evaluating retraining, unlearning, and alternative approaches
- Verifying and auditing machine unlearning without compromising model utility
- Challenges of removing personal, copyrighted, or sensitive information from foundation models
- Emerging benchmarks, standards, and testing methodologies for machine unlearning
- Preparing enterprises for AI governance, compliance, and privacy-by-design in the age of large-scale AI systems
