Welcome to this recap of the second session of The Digital Shelf: Publishing & Library Forum. If you’re new to these sessions, this series is a collaboration between Lyrasis, ReadersFirst, and the Chief Officers of State Library Agencies (COSLA). We launched this forum because we believe that access to knowledge is critical for a functioning democracy, and in a landscape moving this fast, we need a space for honest, practical, and sometimes difficult conversations.

In our latest session, we moved from the broader advocacy updates of our inaugural "Action Round Up" to a deep dive into one of the most pressing issues on the "shelf" today: Ethical AI in Libraries.

The conversation was led by Micah May (Lyrasis), with a special welcome from Michael Blackwell (St. Mary’s County Library/ReadersFirst) and Jeremy Johannesen (COSLA). They were joined by a panel that really knows the "plumbing" and the philosophy of library tech: Marshall Breeding, Peter Musser from ISKME, and Michael Hanegan, author of Generative AI and Libraries.

Here are some highlights from the conversation.

Moving Beyond "Because We Can"

The tone was set early by Marshall Breeding, who reminded us that just because a technology is available doesn't mean it's ready for prime time in a library setting. There is a fundamental tension between the "Wild West" of generative AI and the library brand of objective, verifiable information. Marshall’s takeaway was clear: libraries shouldn’t just embrace these tools because they’re flashy; we have to ensure there are guardrails in place that protect the integrity of the information we provide to our communities.

The Truth About Algorithmic Bias

Peter Musser, who leads the ALA Core’s AI & Machine Learning in Libraries interest group, pulled back the curtain on the invisible "party lines" baked into these models. Whether it’s a model like DeepSeek, which is trained to only provide answers in line with the Chinese government’s stances, or Western models that reflect a narrow cultural dataset, bias is pervasive. The panel discussed how we, as library workers, have to be the ones to teach our patrons to interrogate the machine. We have to help people understand that these aren't infallible truth-engines; they are statistical models whose outputs are always worth verifying.

Leveraging the "Library Advantage"

One of the most inspiring parts of the talk came from Michael Hanegan, who challenged the idea that libraries are just passive observers in the AI revolution. While trust in government and traditional media has plummeted, public trust in libraries remains incredibly high—around 80-85%. Michael’s provocation was that we should "leverage this to the hilt." We aren't just "nice people who like books"; we are information scientists. We have the skills to move the world toward a more ethical future by scaling up the information literacy work we’ve already been doing for decades.

Specificity Over Outrage

We couldn't talk about AI without touching on the environmental cost. We’ve all seen the headlines about the massive water and energy usage of data centers, but the panel urged us to lead with specificity. As Michael noted, not all players are the same — some companies are transparently working toward "water positive" goals, while others operate with far less accountability. As librarians, our role is to demand that data and hold providers to the same standards we expect for any other part of the publishing ecosystem.

A Shared Effort

This forum is a shared space because these are shared challenges. Whether it's the work Michael Blackwell is doing at ReadersFirst to ensure fair digital licensing or Jeremy Johannesen’s leadership at COSLA in supporting state-level library health, the "digital shelf" only works when we work together.

What’s next? As we look ahead into 2026, we’ll be continuing these conversations with sessions on Sustainable Pricing for Ebooks and the Banned Book Index Project. We want to keep finding actionable, collaborative strategies that serve the public good.

  • Watch the Ethical AI in Libraries Replay: Zoom Recording
  • Catch up on the series: Read the full announcement here.
  • Continuing Education: If you’re a librarian, don’t forget that participating in these sessions makes you eligible for a continuing education certificate from Lyrasis Learning!
  • Up Next: Please join the next Digital Shelf session, which will address how we can work toward Sustainable Pricing for Library Ebooks and Audiobooks. 

Unanswered question from the webinar

I work with a region of 24 small to mid-size libraries. A handful use tools that integrate AI, such as for email composition, editing, grant writing, marketing etc. However, it doesn't seem they have encountered a lot of patron questions about AI or integration in their automation systems. What kind of timeline do you currently see with integration of AI into ILS for public libraries, and do you see any kinds of tools that might be unexpected, beyond cataloging assistance, discovery, etc.?

Marshall Breeding: Compared to academic and research libraries, public libraries generally show a lower level of interest and more skepticism regarding integrating AI technologies into their automation systems.  It’s not that these libraries are necessarily resistant to change, but there have not been many use cases that prove the  value of these technologies to make their internal workflows more efficient, such as the processing of materials or circulation tasks.  The tools for access and discovery appreciated by public libraries are likewise different than academic libraries.  Their collections are more finite and are fairly well handled current technologies.  While it seems likely that the providers of automation systems and discovery services for public libraries will incrementally include some AI-based components, it’s hard to see that these will bring drastic improvements.  AI technologies could help public libraries improve their outreach and patron engagement efforts, though such processes would have to be executed quite carefully to ensure positive results.   

It’s also important to note that many small and mid-sized public libraries may not have access to advanced systems and technologies and may rely on systems acquired a decade or so ago.  A surprising portion of small public libraries have no automation systems at all.

Will introducing AI technologies in to public library automation systems reduce their cost? 

It doesn’t seem likely to me.  Introducing new capabilities, probably with higher costs, will benefit the larger tier of public libraries, though I’m skeptical that it will reach into the realm of smaller public libraries that struggle to afford more basic systems.