Governing AI With Trias Politica

This master’s thesis investigates how the principle of Trias Politica—the democratic separation of powers—can be adapted as a governance framework for artificial intelligence (AI) systems in organizational settings. Building on the proposal by van Gils & Mulder (2023), the study explores how a structured model of checks and balances can help mitigate bias-related risks and enhance accountability in AI implementations. The research is grounded in three key data sources: a structured literature review, a case study of Achmea’s MASHA AI system, and insights from expert workshops and focus groups. Through a deductive thematic analysis, six critical governance factors are identified: Ethical AI principles Separation-of-powers structure Lifecycle governance via policy Procedural and technical safeguards Stakeholder inclusion Alignment with regulatory frameworks (e.g., the EU AI Act) An expert validation panel confirmed the relevance and completeness of the model, offering refinements to strengthen its applicability. The findings suggest that while algorithmic bias in AI cannot be entirely eliminated, it can be effectively managed through transparent, structured, and accountable governance mechanisms. This research contributes to the growing field of AI governance by offering a democratically inspired, actionable model for organizations seeking to deploy fair, trustworthy, and regulation-compliant AI systems.