AI Must Not Become the Curriculum's Invisible Canon

A learner asks an artificial-intelligence system for the best explanation of a public policy, a scientific question or a biblical text. The answer is coherent, confident and conveniently arranged. It may even provide citations. Yet a prior educational question remains hidden: who decided which sources were worth retrieving, which language carried the question faithfully, which voices were ranked first and which forms of knowledge disappeared before the answer reached the screen?

Curriculum has always selected knowledge. Teachers, examination bodies, publishers, churches and professional communities decide what learners encounter and what counts as authoritative evidence. Artificial intelligence does not remove that selection. It adds another powerful layer whose training data, retrieval systems, ranking rules and summaries are often invisible to the learner and only partly visible to the institution.

When that layer becomes routine, AI can function as an invisible canon: not a formally approved list of authoritative texts, but a practical system for deciding what is seen, repeated, trusted and forgotten. Education should use AI's ability to connect knowledge while refusing to let convenience become unaccountable authority.

UNESCO has scheduled a virtual roundtable for 2 September titled Plural Intelligences: Designing Agentic AI that Respects Cultural and Epistemic Diversity. Hosted through its Business Council for Ethics of AI with the Rome Call for AI Ethics, the session asks how agentic systems may reshape the creation, validation,…