A Protocol to Protect Cultural Collections in the Age of AI

One of the bedrock principles of Rare Book School is meticulous care and vigilance for the integrity of the human record. We encourage rigor in methods and in ethics practiced by libraries, special collections, archives, museums, and in the book trade. Our shared vision focuses not just on finding answers in books, but in communicating the practice by which we arrive at our inferences, including transparency about the sources and methods we use.

A threat of the new age of AI is that some models can deracinate information from its sources, effectively removing the necessity to “show the work” and any responsibilities to the repositories from which data has been drawn.

It is for this reason that RBS endorses the University of Virginia Archival AI Protocol, developed by University Librarian Leo S. Lo, which makes a case for the protection of unique cultural collections in light of the new age of AI.

One simple, powerful, and non-negotiable principle of this protocol is “No access without institutional control.” Just as we hold ourselves to high standards as scholars, researchers, librarians, and curators, the protocol argues we should guard against AI models that do not regard the provenance of information and do not allow institutional control to halt improper uses of information.

The University of Virginia is leading the way in how institutions can act together to address this issue and shape the future of AI. To understand more about this protocol, we recommend the following links to our global community of practice.

Read the full protocol here.

Watch the 7-minute video overview of the protocol below: