PETs and AI: Privacy Washing and the Need for a PETs Evaluation Framework (Dagstuhl Seminar 25112)

Abstract

As public awareness of data collection practices and regulatory frameworks grows, privacy-enhancing technologies (PETs) have emerged as a promising approach to reconciling data utility with individual privacy rights. PETs underpin privacy-preserving machine learning (PPML), integrating tools like differential privacy, homomorphic encryption, and secure multiparty computation to safeguard data throughout the AI lifecycle. However, despite significant technical progress, PETs face critical policy and governance challenges. Recent works have raised concerns about efficacy and deployment of PETs, observing that fundamental rights of people are continually being harmed, including, paradoxically, privacy. PETs have been used in surveillance applications and as a privacy washing tool. Current approaches often fail to address broader harms beyond data protection, highlighting the need for a more comprehensive privacy evaluation framework. This Dagstuhl Seminar brought together scholars in computer science and law, along with policymakers, regulators, and industry leaders, to discuss privacy washing and the challenges of detecting privacy washing through PETs and explored pathways toward a framework to address these challenges.