AI Language Tools Risk Replicating Colonial Extraction Without Indigenous Data Sovereignty
Original framing: “Project Harnesses AI To Revive Indigenous Languages” — bing news
Indigenous scholars (e.g., Dr. Keolu Fox, Dr. Kim TallBear) warn that AI trained on scraped or 'donated' language data separates knowledge from its relational context — elders, land, ceremony, and accountability. The SFU project's framing of 'harnessing AI' inverts the proper relationship: technology should be harnessed by communities, not communities by technology. Without free, prior, and informed consent (FPIC) embedded in model architecture, this replicates the extractive logic of salvage anthropology.
The SFU project exemplifies techno-solutionism's colonial grammar: it frames language loss as a technical deficit solvable by external innovation, obscuring that loss was caused by deliberate state policy (residential schools, Sixties Scoop, land theft).
The 8 dimensions converge on one truth — revitalization is political, not computational. Te Hiku Media in Aotearoa proves that when Māori control the data, the models, and the deployment, AI serves sovereignty. When universities control them, AI becomes high-tech salvage anthropology. The trickster watches: the tool that promises to save the language may become the archive that replaces the speaker. The only path forward is the one Indigenous leaders have named for decades: land back, language back, data back — in that order.