MPI-SP Research Featured at CNIL's 2026 Privacy Research Day
The 5th edition of the Privacy Research Day was held on June 24th, 2026 by CNIL (Commission nationale de l'informatique et des libertés—the French data protection agency). The Privacy Research Day is an interdisciplinary event that brings academics, regulators, and practitioners together to address important questions at the intersection of personal data protection, privacy, and AI regulation. This year’s event is unique in that it is organized alongside the G7 Data Protection Authorities meeting. As such, the attendants include data protection authorities (DPAs) from Canada, France, Germany, Italy, Japan, the United Kingdom, and the United States.
MPI-SP researchers gave four presentations at this year’s Privacy Research Day. One of them was further honored with the CNIL-EHESS Award, given jointly by CNIL and EHESS (School for Advanced Studies in the Social Sciences) in recognition of interdisciplinary research that addresses data protection and civil liberties.
Gabriel Lima (Doctoral Researcher, Human-Centered Security and Privacy) presented “Do Citizens Agree with the EU AI Act? Public Perspectives on Risk and Regulation of AI Systems.” The paper was published at the Proceedings of 2026 CHI Conference on Human Factors in Computing Systems (ACM CHI 2026) and received the CNIL-EHESS Award. This study identified that the EU AI Act—the first comprehensive regulatory framework for artificial intelligence (AI)—is misaligned with the general public’s expectations of AI regulation. In contrast to the AI Act’s approach that prohibits, regulates, or exempts AI systems based on the risk they pose to fundamental rights and other protected values, the study found that people from Germany, France, Spain, and the US support policies that regulate all AI systems without outright prohibitions. This project is an interdisciplinary collaboration between human-computer interaction researchers at MPI-SP and legal scholars at the Karlsruhe Institute of Technology (KIT), demonstrating the power of such collaborations to inform AI policymaking.
Ankolika De (Research Intern in 2025, Human-Centered Security and Privacy) presented “What is Safety? Corporate Discourse, Power, and the Politics of Generative AI Safety,” also published at ACM CHI 2026. This study examined corporate discourse on AI safety by analyzing what AI companies focus on when discussing the topic and how they talk about it. By analyzing public-facing statements from Google, Anthropic, and OpenAI, the study identified how these companies construct their own authority in AI safety, describing the risks posed by AI through metaphors with other high-risk technologies. The findings call for broadening AI literacy efforts to include critical analysis of corporate claims, ensuring that debates on AI safety are not dominated by AI companies.
Tamalika Mukherjee (Research Group Leader, Privacy and AI) presented “Equitable Differential Privacy,” published in Frontiers in Big Data. The paper examines how differential privacy (an emerging privacy-enhancing technology) can protect sensitive data while still creating equity concerns. The study broadens the discussion by arguing that equitable outcomes depend not only on technical design but also on inclusive communication throughout the deployment of a differentially private system. It defines “equitable differential privacy” as the design, communication, and implementation of privacy-protecting systems, and draws on Inclusive Science Communication to identify more accessible engagement strategies. Using the 2020 U.S. Census as a case study, the paper offers lessons for government organizations adopting differential privacy in public data systems.
Yixin Zou (Research Group Leader, Human-Centered Security and Privacy) presented “Auntie, Please Don’t Fall for Those Smooth Talkers: How Chinese Younger Family Members Safeguard Seniors from Online Fraud,” published at ACM CHI 2025. The study draws on posts from RedNote, a popular Chinese social media platform, to examine how younger family members step in to protect older relatives from online scams. Their efforts span from financial assistance, emotional support, to education, and, in some cases, more direct intervention. Yet these protective efforts come with real obstacles: older adults sometimes refuse to get help; dark patterns in digital payment systems exploit users with low digital literacy or early cognitive decline; the process for reporting fraud is hard to navigate. The paper calls for closing the gaps in younger family members’ interactions with law enforcement, online platforms, and victim support organizations to better empower them to act as a line of defense against elderly-targeting fraud.