Protecting Learners in AI Enabled Education in Tanzania A Scoping Review of Child Data Governance Algorithmic Bias and Digital Safeguarding

Abstract

Artificial intelligence is expanding across education, but a central problem remains insufficiently examined: AI-enabled learning systems can collect prompts, identifiers, performance traces, device information, voice, images and behavioural signals. In Tanzania, adoption is advancing faster than independent evidence on how children’s data are minimised, stored, shared, retained and contested. The practical problem is therefore not simply whether schools can access AI, but whether learners can benefit without surrendering rights, experiencing discriminatory outputs, or being exposed to opaque commercial systems. This scoping review mapped peer-reviewed and authoritative evidence published principally from 2015 to September 2026. The review followed Arksey and O’Malley’s framework, updated JBI guidance and PRISMA-ScR reporting principles. Searches combined artificial intelligence and educational technology terms with topic-specific, inclusion and Tanzania or sub-Saharan Africa terms. Eligible sources addressed formal education and provided empirical, review, technical or governance evidence relevant to the problem. Evidence was charted by setting, population, technology, method, risk or opportunity, equity dimension and proposed response. Six themes were identified: Data collection and consent, Privacy security and retention, Bias representation and language, Harmful or inaccurate outputs, Procurement and accountability, Safeguarding literacy and remedy. The synthesis showed that technical availability alone is an inadequate indicator of responsible adoption. Evidence from Africa and Tanzania remains limited, fragmented and rarely disaggregated. The paper proposes an evidence-to-action framework connecting context, system practices, learner experience, outcomes, governance and remedy. It recommends staged adoption, documented human oversight, transparent use rules, inclusive alternatives, routine audit and locally grounded evaluation.