<p>Artificial intelligence (AI) is rapidly reshaping how knowledge is created, governed, and applied, yet its development often perpetuates colonial patterns of extraction and misrepresentation for Indigenous communities, leading to what is termed ‘digital colonialism’. This paper critically examines the risks AI poses to Indigenous Knowledge Systems (IKS), including epistemic erasure, data colonialism, and the reinforcement of structural inequities through biased algorithms and exclusionary design, often rooted in an extractive approach to Indigenous data. Drawing on global case studies and the latest scholarship in Indigenous data sovereignty, we argue that superficial inclusion and token representation in AI systems are insufficient and may even exacerbate historical injustices. In response, we introduce the Reciprocal AI Framework—a novel, actionable approach that centers reciprocity, community governance, and mutual benefit in AI design and deployment. This framework operationalizes Indigenous data sovereignty by embedding free, prior, and informed consent, community-led co-design, and culturally aligned outputs into every stage of the AI lifecycle. We illustrate the framework’s potential through real-world examples, such as Māori-led voice AI projects, Aboriginal fire management collaborations, and Indigenous-controlled health data systems, demonstrating how Indigenous leadership can produce more ethical, accurate, and contextually relevant technologies. We conclude that a shift from technical inclusion to structural justice is essential: AI must be reimagined as a tool for Indigenous empowerment, Knowledge Justice, and plural futures. By following the Reciprocal AI Framework, researchers, technologists, and policymakers can help ensure that AI development upholds Indigenous rights, supports self-determination, and enriches the global digital landscape.</p>