Extractive AI as settler‑colonial continuity
Bronte Chiang
Artificial intelligence (AI), and particularly large‑scale generative AI systems, are often framed as neutral or innovative technologies. Critical Indigenous scholars, however, emphasize that these systems operate within ongoing settler colonial structures rather than outside them. From this perspective, extractive AI can be understood as an extension of colonial practices that appropriates land, labour, culture, and knowledge for capitalistic gain. Such extraction is not accidental or metaphorical, but structural and continuous (Tuck & Yang, 2012).
Dr. David Gaertner (2024) at UBC situates AI within a broader settler colonial impulse to enclose and commodify not only physical resources, but Indigenous identity, knowledge, and interior life. Drawing on speculative fiction and Indigenous studies, Gaertner argues that extraction increasingly targets the intangible: stories, language, memory, and futurity. In this framing, AI systems that ingest and reproduce Indigenous texts and cultural expressions without consent or accountability reflect extractive logics.
Indigenous Knowledges fundamentally resist such extraction. They are relational, place‑based, and embedded within networks of responsibility that include land, community, and more‑than‑human relations. Colonial research traditions treat Indigenous knowledge as detachable information, severed from governance and relationship (Smith, 2012). When Indigenous Knowledges are rendered as “data” for AI systems, they are stripped of the contexts and authorities that give them meaning. Chapter Two emphasizes that respectful engagement with Indigenous Knowledges requires attention to authorship, ownership, and responsibility, rather than assuming that access alone confers permission.
When Indigenous Knowledges become “data”: Context, authority, and dispossession
Generative AI systems intensify these tensions through their reliance on massive datasets assembled via large‑scale scraping of digital content. Indigenous stories, language materials, artwork, oral histories, and scholarly writing, often shared for specific cultural, educational, or community purposes, are absorbed into training data without consent, attribution, or benefit to originating communities. This mirrors earlier extractive research practices in which Indigenous Knowledges were collected, recontextualized, and circulated for institutional gain under the guise of innovation and progress (Smith, 2012). At the same time, AI systems can generate content that appears Indigenous in form or style without any lived experience, community accountability, or relational authority, further complicating questions of who speaks for Indigenous knowledge (Lewis, 2020).
From an information literacy perspective, extractive AI demands critical interrogation rather than technical mitigation. Rather than resolving long‑standing information harms, extractive AI intensifies them: datasets that privilege dominant voices, algorithms that determine relevance through settler logics, search systems that strip knowledge of context, descriptive practices that enact epistemic harm, and copyright rules that enable ongoing dispossession. These conditions are not fully addressed through improved design alone, and acknowledging this limitation is essential for ethical engagement with AI in contexts involving Indigenous Knowledges.
