The Indian Community School (ICS), an intertribal school in southeastern Wisconsin, has implemented an early-grade artificial intelligence literacy progression designed to teach how machine learning systems function before students are granted direct access to AI platforms. Starting in the second grade, learners study pattern recognition, training data, algorithmic bias, and system hallucinations through unplugged classroom exercises. In one foundational activity, students sort physical buttons by attributes such as size, shape, and colour to simulate human data labelling, identifying ambiguous items that illustrate how inconsistent inputs compromise model accuracy.

As students advance into middle school, the curriculum broadens to examine data sovereignty, user privacy, and the physical and environmental costs associated with running AI infrastructure. Controlled access to generative tools begins only in the sixth grade. Rather than receiving unrestricted accounts, students must articulate a specific educational rationale—such as requesting writing feedback, brainstorming, research, or image generation—which teachers evaluate before granting temporary access for that specific task.

The ICS model offers an alternative to typical district responses, which have often focused narrowly on establishing acceptable-use policies or banning commercial chatbots after the fact. By establishing an understanding of data mechanics, categorisation, and ethics through tactile materials first, the programme treats AI literacy as a foundational computer science discipline rather than mere software training.

For South African schools delivering the Department of Basic Education's Coding and Robotics curriculum, this approach illustrates how early machine learning concepts can be introduced via unplugged activities without relying on expensive hardware or high-bandwidth connectivity.