Indigenous AI Builders Embed Data Sovereignty Into Platforms

"[Data] determines what the system sees, what it optimizes for, who benefits, and who is harmed," Tsosie said. "Our community's data is already on this pathway, and the question is whether or not our community governs what it becomes."

Tsosie called for research models where samples and data stay on ancestral lands and research questions come from community members. Community benefit, she said, should be the design goal, not an afterthought.

Canada has a long history of federal misuse of Indigenous data for surveillance and erasure. That led to standards such as the First Nations Principles of OCAP (ownership, control, access, possession), which govern how data can be used. At the Mila event, AI and digital innovation minister Evan Solomon acknowledged the "scar tissue" between the government and Indigenous communities on data use. Solomon was gifted a braid of sweetgrass by University of Calgary doctoral candidate and Mila fellow Mick Elliott, who said it carried a "great weight and responsibility" toward Indigenous people.

Michael Running Wolf, co-founder of Mila's First Languages AI Reality (FLAIR) program, acknowledged that building AI might seem unattractive given how large tech companies have used the technology for exploitative purposes. The goal of having more Indigenous representation in AI, he told the audience, "isn't to contort ourselves, so that we break ourselves."

"It's that we demonstrate the flexibility of our culture and the behaviour as representatives of our culture," he said.