Do Artificial Agents Need Privacy?

Name: Mireille Hildebrandt
Organisation: Professor of Smart Environments, Data Protection and the Rule of Law, Radboud University Nijmegen and Institute of Computing & Informationa Sciences (ICIS), Vrije Universiteit Brussel

Artificial Intelligence (AI) has grown beyond science fictional wishful thinking. Much of our online and part of our offline environments are constituted by computational infrastructures that are both artificial and intelligent. Search engines, behavioral advertising, high frequency trading, knowledge management systems, remote healthcare often thrive on AI in the sense of machine learning or proactive computing. All this AI in turn thrives on aggregation of data and subsequent pattern recognition that is then applied to individual ‘users’ who may find their privacy compromised. Much has been written about this. Today I want to turn the tables and raise the question whether – and if so, under what conditions – artificial agents would qualify as requiring a measure of privacy. By confronting this counterintuitive question I hope to clarify (1) what is meant with an (artificial) agent and (2) what we mean when we speak of privacy.