AI and our interviews
The CSUSM University Library Special Collections department collects, curates, and provides access to resources of historical, cultural, and scholarly importance and value. We are committed to making these collections available for research, teaching, and public engagement while protecting the rights and expectations of donors, creators, custodians, and communities. These obligations extend to those that participate in our oral histories; narrators, interviewers, and collaborators.
In the age of generative AI, information can circulate at scale without clear origin, context, or accountability. For oral histories, this can create irreversible risks to privacy, community interests, donor intent, and the integrity of the public record. Because AI model training is effectively irreversible, we do not permit generative AI training on our oral history materials. This includes researchers who wish to ingest AV files and transcripts of oral histories that are stewarded by CSUSM Special Collections, University Archives, and Scholarly Communications for data mining in generative AI platforms.
We may support AI projects that use retrieval over our resources or that build controlled tools and models under our oversight, when source materials remain under our control, and these projects align with our mission and respect donor and community obligations. All AI-related uses of our resources require prior authorization from the University Library, are pursuant to consent forms, donor and transfer agreements, and may require additional authorization from narrators, interviewers, donors, and transferring university units. To request any generative AI use of archival collections, please contact archives@csusm.edu. We will acknowledge receipt of your request and outline a process for evaluation of your request within ten business days.
This work is based on the UVA Archival AI Protocol, created by Leo S. Lo and developed with the UVA Special Collections and Preservation staff, University of Virginia Library. Canonical URL: https://doi.org/10.18130/5dqf-9w86.