This study presents a proof-of-concept examination of how generative artificial intelligence tools, text-to-speech, text-to-image and text-to-video platforms, interpret translated Ao Naga oral traditions. Using a small set of folk songs and narrative descriptions, the study evaluates the cultural accuracy, semantic coherence and representational fidelity of artificial-intelligence-generated outputs. The findings reveal both the potential of multimodal artificial intelligence technologies for enhancing access to Indigenous cultural materials and their significant limitations, including mispronunciation, cultural flattening and stereotypical visual representations. These outcomes align with broader concerns about digital colonialism and the structural biases embedded in mainstream artificial intelligence training data sets. As the study does not include community participation, the results should be viewed as preliminary technical observations rather than culturally validated interpretations. The work establishes a baseline for future participatory research involving Ao knowledge-holders and highlights the need for culturally grounded and ethically informed approaches to artificial intelligence use in Indigenous heritage contexts.
Supongtemjen et al. (Thu,) studied this question.