Pursuing technology sovereignty under dependency: The case of Mistral AI

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

This study examines how organizations pursue technological sovereignty while embedded in infrastructures controlled by geopolitical competitors. We address this question through an in-depth longitudinal case study of Mistral AI (2023–2025), Europe's only frontier-scale Large Language Model developer competing with American incumbents. Our analysis draws on two primary data sources collected over the firm's first three years: 16 media interviews with CEO Arthur Mensch and 30 monthly activity reports published by Mistral, triangulated with technical documentation, industry analyses, and European policy documents. We employed an abductive analytical approach, combining iterative open and focused coding of institutional work practices with theoretical dimensioning and process tracing across four temporal phases of the company's evolution. Our findings reveal three interrelated forms of institutional work: selective decoupling of critical dependencies while maintaining beneficial connections, strategic leveraging of embeddedness in some structures to contest dominance in others, and structural reconfiguration creating alternative institutional arrangements. These practices constitute geopolitical embedded agency, a distinctive form of organizational action characterized by material rather than symbolic embeddedness; cross-scalar institutional contradictions; and sovereignty objectives that reshape conventional organizational rationality. Our framework contributes to institutional work theory and emerging technology sovereignty scholarship while extending analytically to other capital-intensive sectors facing geopolitical dependencies.