Artificial general intelligence (AGI) represents an unprecedented ambition within the field of technology, aiming to create systems capable of matching or surpassing human abilities across multiple domains. Unlike Artificial Narrow Intelligence (ANI), AGI is anticipated to operate without task-specific limitations and predefined purposes, raising complex, pressing issues surrounding autonomy, control and potential societal impact. This article applies Rittel and Webber’s wicked problem theory to critically examine AGI governance, categorising AGI within the ten characteristics of wicked problems. The absence of a definitive formulation, its unstoppable potential evolution, the subjective and context-dependent nature of its solutions, the irreversibility of interventions and the multiplicity of stakeholder perspectives all underscore the inadequacy of existing governance paradigms. In response, this article advocates for dynamic, iterative and flexible governance frameworks that acknowledge AGI’s ontic uniqueness and potential for autonomous evolution. Rather than treating AGI as a distant or hypothetical concern, this analysis argues for a multidimensional, forward-looking governance model that recognises AGI as an urgent and inherently wicked problem.