Systemic critique of AI’s encroachment on poetic expression: Language tech’s colonial extraction of cultural creativity

Indigenous traditions view poetry as a relational practice tied to land, ancestry, and collective survival, not an individualistic or technical endeavour. AI’s treatment of language as data ignores the sacred and communal dimensions of oral traditions, reducing centuries of cultural transmission to algorithmic inputs. The lack of Indigenous co-authorship or governance in such projects reflects a broader pattern of epistemicide, where marginalised knowledge systems are erased in the name of 'progress.'

The commodification of poetic expression mirrors historical patterns of cultural appropriation, from the colonial theft of indigenous art to the 19th-century European fascination with 'exotic' oral traditions. The 20th-century avant-garde’s fetishisation of 'primitive' art laid the groundwork for today’s Silicon Valley mythos of 'disruptive creativity.' Meanwhile, the printing press’s role in standardising language and marginalising regional dialects offers a parallel to how AI homogenises poetic styles under the guise of 'accessibility.'

Cross-culturally, poetry is often a tool for resistance and decolonial practice, from the *nuyorican* spoken word movement to the Palestinian *zajal* tradition. AI’s depoliticisation of poetic language risks neutralising these subversive functions, turning dissent into 'content.' The Japanese *haiku* tradition, with its emphasis on *kigo* (seasonal references) and *aware* (pathos), contrasts sharply with AI’s lack of emotional or ecological embeddedness. Similarly, the African *oraliture* tradition challenges the written-word bias of Western literary AI.

AI language models are trained on vast datasets scraped from the internet, often without consent or compensation, raising ethical concerns about data sovereignty and copyright. Studies show that these models perpetuate biases present in their training data, including racial, gender, and cultural stereotypes in poetic output. The lack of transparency in dataset composition (e.g., overrepresentation of Western canonical works) skews AI-generated poetry toward Eurocentric forms. Meanwhile, research on human-AI collaboration in creative fields highlights the risk of deskilling and the erosion of critical discourse.

Poetry is not merely a linguistic exercise but a spiritual and artistic act that connects humans to the ineffable, the ancestral, and the ecological. AI’s reduction of poetry to pattern-matching strips away its capacity to evoke awe, grief, or transcendence, reducing it to a hollow simulacrum. The loss of metaphor’s depth—where words carry layers of meaning beyond literal interpretation—is a spiritual impoverishment. Indigenous and mystical traditions (e.g., Kabbalah, Sufism) treat language as a vessel for divine or cosmic truth, a dimension entirely absent in AI-generated verse.

If unchecked, AI-driven poetic tools could lead to a homogenisation of global literary traditions, where 'AI-optimised' styles dominate publishing and education, eroding linguistic diversity. The monetisation of AI poetry risks creating a two-tier system: elite 'human' poets versus mass-produced algorithmic content, further marginalising non-Western voices. Conversely, if governed by ethical frameworks like data sovereignty and communal ownership, AI could be used to revitalise endangered languages or amplify marginalised voices—but this requires structural change, not just technical fixes.

Marginalised poets—especially Black, Indigenous, and Global South creators—are disproportionately affected by AI’s extractive practices, as their works are scraped without credit or compensation. The tech industry’s 'move fast and break things' ethos has historically harmed communities of colour, from redlining algorithms to facial recognition biases. Without reparative justice mechanisms, AI poetry tools risk becoming another tool of cultural erasure, where the profits flow to Silicon Valley while the creators remain unacknowledged.