CRL Publications

Large Scale Digitization in the AI Era

Issues and Model Terms for Libraries and Their Partners

Libraries are increasingly being approached about partnerships that would digitize collections at scale and potentially make the resulting digital content and metadata available for commercial artificial intelligence or other proprietary services. These proposals may offer significant benefits—including expanded access, preservation, and reduced digitization costs—but they also raise questions about ownership, permitted uses, public access, institutional control, compensation, attribution, privacy, and long-term stewardship.

CRL member library directors have identified a need for clearer guidance about the terms libraries should seek when considering these arrangements. Decisions made by individual institutions may also shape emerging practices across the library community and affect how collectively stewarded scholarly and cultural resources are used.

About the Issue Paper

This briefing document was developed in response to discussion among CRL member library directors about the terms that libraries should seek when collections are proposed for large-scale digitization, particularly where the resulting digital content and metadata may support commercial AI or other proprietary services.

This public version is being released to support all libraries in navigating these issues, which CRL members see as a matter of broad professional concern. CRL’s primary role at this time is to help libraries make informed decisions by gathering experience and understanding member questions in order to develop baseline terms for partnership agreements. 

CRL intends for this to lead to practical action, but the particular courses of action to be taken—from negotiations over any specific partnership, pooled requests for proposals or similar means of brokering commercial use of library content, or developing a shared corpus under specific model terms—require further member discussion and legal review.

Read the Paper: