Predictive risk modelling using administrative data is increasingly being promoted to tackle complex social policy issues, including the risk of child maltreatment and recurring involvement with child protection systems. This paper discusses opportunities and risks concerning predictive risk modelling with administrative datasets to address Indigenous Australian overrepresentation in Australian child protection systems. A scoping review using five databases, and the Google search engine, examined peer‐reviewed and grey literature on risks associated with predictive risk models (PRMs) for racial and ethnic populations in child protection systems, such as Indigenous Australians. The findings revealed a dearth of research, especially considering Indigenous populations. Although PRMs have been developed for Australian child protection systems, no empirical research was found in relation to Indigenous Australians. The implications for utilising administrative data to address Indigenous Australian overrepresentation are discussed, focusing on methodological limitations of predictive analytics, and notions of fairness and bias. Participatory model development, transparency and Indigenous data sovereignty are crucial to ensure the development of fair and unbiased PRMs in Australian child protection systems. Yet, while PRMs may offer substantial benefits as decision support tools, significant developments – which fully include Indigenous Australians – are needed before they can be used with Indigenous Australians.