The Colgate University Libraries have joined an inaugural cohort of 19 university libraries in signing a Statement of Shared Practice on AI and Archives. The framework guides how libraries evaluate requests from AI companies seeking access to archival and special collections for model training. Colgate is the first liberal arts institution to join the cohort, which is coordinated by the University of Virginia Libraries.
The statement establishes a set of professional standards for how cultural institutions assess AI training requests while protecting provenance and maintaining commitments to those who contribute to and are represented by the collections. This is outlined through six shared commitments that provide a collective framework and a support network for archivists by advocating for shared definitions and transparency.
“Archivists carry an explicit duty of care toward donors, records creators, and communities regarding the specific terms under which their materials were deposited,” says Joshua Finnell, associate university librarian and director of research and scholarly initiatives. “This emphasis on documented provenance ensures that future researchers can interrogate and build upon historical evidence with a comprehensive understanding of its creation.”
Provenance refers to the documented history of a record’s creation, ownership, and preservation — a central principle of archival practice. Library leaders involved in the initiative argue that current AI training models can weaken those connections by incorporating archival materials into systems that may not preserve clear links to their original sources.
This is particularly relevant for Colgate’s Special Collections and University Archives, which house unique materials including the Colgate Family Papers — documents that catalog the history of the University’s namesake and the founding of the Colgate-Palmolive Company. The archives are also home to the Richard S. Weiner Collection of George Bernard Shaw, consisting of more than 1,500 letters, first editions, manuscripts, and other writings.
Each month, representatives from participating institutions meet to discuss how their libraries are evaluating, negotiating, and governing AI partnerships while honoring obligations to donors, their communities, and the integrity of the public record.
“These shared commitments aim to encourage intentional and responsible engagement with AI training models, and seek to protect the deep trust imbued in these cultural institutions,” says Finnell.