Public-Sector Adoption of Privacy-Preserving Machine Learning

Abstract

Background: Policy attention to privacy-preserving ML has increased, yet there is limited empirical clarity on how institutional arrangements influence the public sector and implementation quality.

Objective: The study evaluated institutional variation in approaches to privacy-preserving ML, with particular attention to accountability, administrative capacity, and the public sector.

Methods: Survey data were collected from 154 relevant officials, professionals, service users, or regulated actors using measures of implementation quality, capacity, procedural clarity, transparency, and perceived effectiveness. Descriptive and adjusted regression analyses identified correlates of stronger implementation.

Results: Analysis of 139 complete questionnaires on privacy-preserving ML showed that 61% of respondents experienced consistent implementation, while 12% encountered substantial procedural or capacity constraints. Adjusted models linked better implementation to role clarity, staff capability, and availability of formal guidance.

Conclusion: For privacy-preserving ML, the main practical lesson is that formal requirements do not by themselves produce consistent implementation. This means that responsible AI adoption requires governance arrangements that can be implemented consistently across agencies, not merely high-level principles.

Study design: technology adoption survey.

Publication history: Originally published in print on 15 June 2026, Volume 6, Article 17. Online abstract record added on 16 September 2026. Full-text PDF will be added when the digital galley is ready.