Abstract

In a landscape where governments are shaped by and depend on private AI providers, what does it mean to govern artificial intelligence (AI)? Public procurement is a point of intervention where the entrepreneurial pull of states to integrate digital expertise and reformulate its problem within machine-learning logics can be halted, questioned and examined. This thesis examines public procurement of AI as a crucial site where governments and AI providers engage in a complex co-shaping process, which I term the procurement circuit. Specifically, the thesis examines Canada’s procurement of AI as part of its national Responsible AI Strategy.

Through situational analysis, this thesis maps and explains how this co-shaping occurs and considers how the procurement circuit distributes authority and legitimacy over normative questions on AI between AI providers and government. I argue that Canada's regulatory architecture is built under what Louise Amoore coined Machine Learning (ML) political order. Chapter 3 maps the regulatory architecture Canada built to enforce Responsible AI and evaluate suppliers. Chapter 4 considers 11 suppliers’ responses to these requirements and outlines their normative views on both AI and its governance. In the conclusion, I suggest recommendations on how Canada might reformulate the procurement circuit as a space where legitimacy and authority is negotiated to resist ML political order.