From exclusion to openness: an ethical stage-gate model for reflecting on digital sovereignty in AI systems

Artificial intelligence (AI) is ubiquitous—but not everyone understands how it works or its influence on future developments. While a small group actively trains AI, the majority uses it passively and uncritically. This creates structural asymmetries: AI systems learn primarily from active users, resulting in biased representations and possible exclusion. Groups that only consume AI run the risk of their needs no longer being taken into account in future system designs. This article applies the concept of the stage-gate model (as developed for ethical reflection in the context of selective demarketing) to AI systems and shows how algorithmic processes can reinforce social inequality. The proposed model shows how potential exclusions can be identified at an early stage. The article thus advocates openness—in the sense of open science, open data and open source—as a structural counter-strategy to closed, elitist AI developments. The aim is to systematically integrate ethical reflection into technical design processes.