Reimagining AI in Latin America: situated narratives of users, developers, and decision-makers on understanding and governing AI

Abstract AI technologies are rapidly transforming societies, yet dominant narratives about AI development and governance often emerge from the Global North, overlooking perspectives from Latin America. This study explores how stakeholders in Chile—end-users, developers, and decision-makers—understand and engage with AI, focusing on transparency and governance. We uncover how local narratives shape AI perceptions, understandings, and demands for governance approaches by drawing from qualitative research projects, including interviews, focus groups, and a digital ethnography of AI regulation discussions in the Congress. A key finding is the widespread perception of Chile as a "test subject" or "sandbox nation" for AI across the different stakeholders, reflecting broader power asymmetries in technology development. In addition, stakeholders conceptualize and demand different governance approaches depending on whether they refer to AI as an abstract idea or a specific AI system. End-users often view AI as an opaque force imposed upon them, leading to trust based on functional efficiency rather than informed understanding—however, their demands for transparency and human oversight increase when engaging with specific AI-mediated systems. Developers seek to demystify AI by shifting the focus from abstract capabilities to particular applications, emphasizing its mechanical nature and countering misconceptions shaped by anthropomorphic narratives. Decision-makers, meanwhile, negotiate the definition of AI technologies balancing between broad and specific definitions of AI and navigate tensions over governance frameworks from the Global North (US vs. Europe). This study emphasizes the need for transparency and public participation tailored to local sociopolitical contexts and contributes to global debates on AI’s understanding and governance, advocating for policies that reflect regional priorities rather than universalized frameworks.