The University of Virginia Archival AI Protocol

Abstract

The University of Virginia Archival AI Protocol (UVA AAIP) establishes a practical standard governing how artificial intelligence systems may access and use archival collections. The Protocol distinguishes AI foundation model training, a stewardship decision that permanently changes an institution's relationship with its collections, from retrieval-based AI systems, which keep source materials under organizational control and preserve the provenance chain. Built on three foundational pillars, provenance and attribution, donor and community responsibilities, and institutional accountability, the Protocol conditions access to archival materials on demonstrated item-level provenance, meaningful attribution, and contractually enforceable institutional accountability, applying a presumption against approval for broad commercial training where those conditions cannot be met. The Protocol provides a decision framework, model contract clauses for deeds of gift and vendor agreements, minimum provenance standards for AI-generated citations, and a phased implementation plan. Designed for adoption by memory institutions of any size and type, it offers a consistent position from which to evaluate, negotiate, and govern AI partnerships while honoring obligations to donors, communities, and the integrity of the public record.