Evidencing AI Literacy Measures: An Operational Model for Article 4 of the EU AI Act

Article 4 of Regulation (EU) 2024/1689 was replaced on 27 July 2026 by Regulation (EU) 2026/1744. Providers and deployers must now “take measures to support the development of AI literacy” of their staff and of others dealing with AI systems on their behalf; the obligation expressly “does not require providers or deployers to guarantee any specific level of AI literacy of any individual.” The earlier formulation — measures to “ensure, to their best extent, a sufficient level” — no longer states the law. The practical question therefore moves from proving an abstract individual standard towards showing what an organisation did, for which systems and roles, in which context, and with what review. This Research Note asks what evidence an organisation can retain to show that such measures were designed, implemented and maintained. It fixes the current legal text and separates it throughout from guidance, standards, good practice and the author’s own proposals. It argues that the workable unit of analysis is not the course but the triple system + role + context, and develops an author-proposed framework, the AI Literacy Evidence Chain, linking system, role, context, literacy need, measure, evidence and review. It adds a taxonomy of measures, a ten-category evidence architecture, a lightweight SME record, a fifteen-question self-assessment, and an analysis of recurrent failure modes. No documentation model is presented as legally mandatory; none currently is.