Indigenous artists across Latin America are experimenting with Midjourney and other generative AI tools to imagine their own futures, but they're also discovering that these technologies often distort and misrepresent Indigenous cultures in troubling ways. A new book published by the University of London Press documents the AIAI project, a collaborative research initiative that invited Indigenous artists and writers from Argentina, Bolivia, Brazil, Chile, and Peru to critically engage with image generation technologies between 2023 and 2026.

What Happens When AI Tries to Generate Indigenous Art?

The AIAI project brought together a diverse group of Indigenous creators to explore both the creative potential and the serious limitations of tools like Midjourney and DALL-E. Through hands-on experimentation with commercial and custom-built AI systems, these artists discovered a fundamental tension: while AI image generators can help them "dream" and imagine Indigenous futures, the same tools frequently "hallucinate" and distort Indigenous cultures, perpetuating stereotypes and misrepresenting their cosmovisions.

The resulting book, "Artificial Intelligence, Art and Indigeneity: Between Dreams and Hallucinations," is the first book-length study of generative AI's impact on Indigenous visual culture in Latin America. It brings together artists, writers, and scholars using what researchers call a "decolonial framework rooted in Indigenous epistemologies," meaning it centers Indigenous ways of knowing rather than treating them as objects to be studied by outsiders.

How Are Indigenous Artists Using AI Image Generation Tools?

The book is structured in three main sections that reveal how Indigenous creators are approaching these technologies:

• Context and Theory: The first section establishes the historical and cultural background, exploring how generative AI fits into broader conversations about Indigenous representation and cultural sovereignty.
• Creative Reflections: The central section features chapters written by Indigenous artists and writers themselves, published in their original Spanish or Portuguese alongside English translations, showcasing their direct experiences experimenting with image generation tools.
• Generative Critique: The final section offers scholarly analysis of key issues including stereotypes, creativity, censorship, copyright, and how Indigenous communities can develop their own AI systems that reflect their values and knowledge systems.

Contributors include artists like Tadeu Dos Santos (Tadeu Kaingang), who explores how AI can represent Indigenous people and their spiritual traditions, and Mariela Tulián, who reflects on her dreams about working with AI technology. Their essays reveal both the creative possibilities and the frustrations of using tools designed without Indigenous input.

One particularly innovative chapter, "Dragon Dreaming: Notes on Methodology for Collaborative Design of an Indigenous Image-Generation Prototype," documents how researchers Dave Lynch and Sam Hallas worked with Indigenous communities to design a custom AI image generator specifically built to understand Indigenous visual culture, rather than relying on commercial tools trained on data that often misrepresents Indigenous peoples.

Why Do Current AI Image Generators Fail Indigenous Communities?

The book's analysis reveals a critical problem: mainstream AI image generators like Midjourney are trained on vast datasets of images from the internet, which contain centuries of colonial stereotypes, misrepresentations, and distortions of Indigenous peoples. When these tools are asked to generate Indigenous art or imagery, they reproduce these same biases and inaccuracies.

Sandra De Berduccy, one of the book's editors, contributed a chapter titled "A Genealogy of Stereotypes: The Representation of Indigenous Peoples with Generative AI," which traces how these stereotypes become embedded in AI systems and perpetuated through image generation. The book argues that true inclusion in AI development requires far more than simply adding more Indigenous data or refining prompts; it demands that Indigenous peoples participate in building these technologies from the ground up.

"AI fails to grasp their cosmovisions, and that true inclusion requires more than adding data or refining prompts. A living archive of resistance, media appropriation, and indispensable decolonial and technological critique from Latin America," noted Rubria Rocha de Luna, a professor at Tecnológico de Monterrey in Mexico.

Rubria Rocha de Luna, Professor at Tecnológico de Monterrey, Mexico

The book emphasizes Indigenous agency and self-determination in AI development. Rather than treating Indigenous peoples as passive subjects whose cultures are fed into AI systems, the contributors argue that Indigenous communities should lead the creation of knowledge bases and AI tools that reflect their own values, epistemologies, and visual traditions.

What's the Path Forward for Indigenous-Led AI Development?

The book points toward a concrete solution: Indigenous-created knowledge bases and AI systems. One example highlighted in the book is Etnias, a social network created by Karuãn Tenteneady, an Indigenous systems development technologist from Brazil. Etnias brings together a wide range of materials authored by Indigenous people themselves, creating a resource that could eventually train AI systems to better understand and represent Indigenous cultures accurately.

"At a time when much online information about Indigenous Peoples is shaped by non-Indigenous perspectives, initiatives such as Etnias can help ensure that future AI systems more accurately reflect the diversity and reality of Indigenous peoples," explained Karuãn Tenteneady, an Indigenous systems development technologist and creator of the social network Etnias.

Karuãn Tenteneady, Indigenous Systems Development Technologist and Creator of Etnias

The book's publication comes at a critical moment. As tools like Midjourney become increasingly accessible to artists worldwide, the question of who controls these technologies and whose perspectives they encode becomes more urgent. The AIAI project demonstrates that Indigenous artists are not passive consumers of AI tools; they are active critics and innovators working to reshape these technologies in service of their own cultural visions and decolonial futures.

The enhanced open access edition of the book includes Spanish and Portuguese versions of the full text, video interviews with participating artists, and a curated selection of images and animations created during the project. This multilingual approach reflects the book's commitment to centering Indigenous voices and making the research accessible to communities across Latin America, not just English-speaking academic audiences.