Shadow AI at Work: Labour Control and Everyday Resistance

Shadow AI is commonly framed as the unauthorized use of artificial intelligence tools outside approved organizational governance structures. Existing literature has predominantly interpreted this phenomenon through the lenses of cybersecurity, compliance, legal liability, and organizational oversight. This paper argues that such approaches are incomplete because they understate the labour conditions that generate shadow AI in the first place. Drawing on labour process theory, the paper reconceptualizes shadow AI as a complex workplace practice shaped by contested control, hidden skill, everyday resistance, and the redistribution of responsibility. Re-reading shadow AI literature through this lens shows that many studies already document governance drift, informal practices, trust breakdowns, and structural inequalities, even when they frame these mainly as risk or compliance issues. This work reframes shadow AI as a constitutive feature of the contemporary labour process rather than a peripheral compliance failure. It contributes to a conceptual model of how unsanctioned AI use stems from the interaction of work intensification, hidden skill formation, and downwardly distributed responsibility.