Litmus
Project Overview
University
Year
About
Sensitive personal data is increasingly used to train AI, but claims that it is anonymised or protected are difficult to verify. Litmus tests privacy safeguards without requiring a technology provider’s cooperation. It marks private data in a detectable way, then examines generated outputs for statistical evidence that the information was memorised during training. Building on research that exposed weaknesses in Apple’s analytics system, it could help regulators, hospitals and organisations assess privacy protections using measurable evidence rather than relying on providers’ promises.
Team
Rishav Chourasia
Dr. Rishav Chourasia is Head of Product at Betterdata and an AI/ML privacy expert with a PhD in Machine Learning from the National University of Singapore (NUS). His work focuses on privacy-preserving machine learning, differential privacy, and machine unlearning, bridging advanced research with real-world product applications. He has been recognized for breakthrough research in privacy auditing, differential privacy dynamics, and machine unlearning. Before Singapore, he earned his B.Tech in Computer Science from Indian Institute of Technology Guwahati, spent a summer as a student researcher at Google Brain, and held a research fellowship at the Max Planck Institute for Software Systems.