Cedric Whitney

Cedric Whitney

Research Affiliate

Cedric Whitney is a researcher specializing in mixed-methods evaluations of AI governance mechanisms, combining qualitative, quantitative, and adversarial testing approaches. His dissertation work focuses on the limits and boundaries of privacy-enhancing technologies in broader-purpose governance. His dissertation examines how synthetic data and machine unlearning operate in practice—questioning when they meaningfully improve privacy and compliance versus when they introduce new risks. Cedric’s research has been recognized with a Best Paper award at FAccT 2024 for his work on “metric washing,” exposing how synthetic data can distort AI fairness and safety evaluations. He is also a recipient of the NSF Graduate Research Fellowship (NSF GRFP).

Cedric’s current passion project—and the final chapter of his dissertation—investigates alignment stress testing through the lens of contextual integrity. His research explores whether model adherence to privacy and information-sharing principles is genuinely robust or simply optimized for in-distribution compliance, a question of growing importance as AI systems are integrated into agential workflows. Beyond academia, Cedric has applied his expertise in regulatory, industry, and civil society settings, from the FTC and the CPPA to CDT and IBM, and his pre-PhD work involved leading partnerships and deployments of federated machine learning for healthcare AI unicorn Owkin.