article · Journal of Linguistics Literary and Communication Studies
The integration of generative artificial intelligence (AI) into political resistance marks a consequential shift in the mediation of dissent. This paper examines a paradox at the heart of AI music activism: generative audio tools appear to democratise sonic production for marginalised dissidents, yet the surrounding infrastructures operate through logics of cultural extraction, principally the ingestion of uncredited African musical material into foundation-model training corpora and the capture of engagement value by platform distribution algorithms. Drawing on a retrospective qualitative digital ethnography of discourse concerning the 2024 Kenyan #RejectFinanceBill2024 and #OccupyParliament protests, the study analyses how activists and suspected state-aligned actors deployed AI audio systems across X and adjacent networks. The corpus comprises purposively selected creator posts, reception threads and named commentary from X and cross-posted TikTok material, analysed through reflexive thematic analysis and supplemented by computational audio analysis of the disputed Tujadiliane track. The analysis triangulates mass self-communication, networked counterpublics, Oramedia, platform capitalism and data colonialism. Findings show that Kenyan activists engineered what the paper terms Digital Oramedia by prompting generative systems to synthesise Sheng lexicon, isukuti-inflected rhythmic textures and chant-like sonic structures, enabling rapid mobilisation beyond legacy gatekeepers. Audiences simultaneously perceived a countervailing flow of suspected state-aligned material, most prominently the synthetic reggae track Tujadiliane, producing epistemological anxiety alongside outrage at the appropriation of an anti-colonial idiom. Audience discourse indicates a substantive shift in verification norms from source authenticity toward cultural-political alignment. The paper argues that sustainable AI music activism requires decolonial interventions at both the training-data and distribution layers.
This page summarises published work. The authoritative version sits with the publisher.
DOI: 10.58721/6gjgeb50
Is something wrong with this record? Report it or request removal.
Discussion
Have you built on this work, tried to replicate it, or seen it applied in practice? Share what you know. Verified researchers and MARATTO™ domain experts can open a discussion, and any member can reply. Contributions are reviewed before they appear.
No discussion yet. Open the first thread.
New to MARATTO™? Create a free account.
