In a meeting room at the C. V. Starr East Asian Library, a Chinese film poster flickers onto a large screen. Haiqing Lin, the library's head of technical services, is demonstrating an interface he built. With a few practiced clicks, he sends the image—one of 2,000 posters from the Paul Fonoroff collection—through an artificial intelligence model. Seconds later, a draft catalog description appears.
For Lin, who has spent his career thinking about how to describe materials so researchers can find them, the appeal is obvious. Cataloging each poster is slow work. His custom tool, which incorporates Google's Gemini AI models via an application programming interface, can accurately translate the language and generate descriptive metadata to be folded into the UC Berkeley Library's catalog. Though it's still in an exploratory phase, it shows real promise.
An early adopter of AI technologies, Lin has experimented extensively, continually refining his latest prompt. He is convinced that AI will transform cataloging work but only as a tool to support expertise, not replace it.
"To develop this prompt, we need knowledge about the collection, the user's expectations, and the cataloging system," he said. "This is our librarians' expertise. We teach AI what we need. We control AI, rather than it controlling us."
This kind of grounded experimentation is unfolding amid broader uncertainty about AI's role in academic libraries. As the technological revolution accelerates, the Library is taking a deliberate, human-centered approach, weighing the merits and missteps of these tools, testing where they can help, where they fall short, and how to use them responsibly.
Steering that course is Carolyn Caizzi, the associate university librarian for digital initiatives and information technology. And she's bullish on the Library's ability to navigate the moment.
"AI is here, and we have to figure out a path forward," she said. "So it's about adapting and redefining our role. I think we're well-suited to it because we are information professionals. We know the questions to ask."
Unlocking hidden histories
For Becky Miller, the university's natural resources librarian, AI offered a solution to a long-standing challenge. For years, she'd had her eye on shelves of agricultural extension publications, primarily from the 1960s and '70s, that documented everything from crop yields to pest outbreaks. The publications had sat untouched for decades in the Marian Koshland Bioscience, Natural Resources & Public Health Library.
But an opportunity arose when Miller joined the Digital Lifecycle Program working group, led by Lynne Grigsby, the head of Library IT. The team was piloting a new AI-powered metadata tool, JSTOR Seeklight. And they needed real-world test cases.
Running the digitized documents through Seeklight produced searchable metadata for materials that had effectively been invisible to researchers. Miller reviewed the output, and 79 records were loaded into the Library's Digital Collections portal this winter. They were the first items in that system with AI-generated descriptions. There are now more than 300 items in the University of California Agricultural Extension Publications collection.
"It's exciting to think about other things that I've been wanting to have digitized," she said. "If we can finally get them posted in a way that people can find and use them, that is very valuable."
The documents could be useful for scientists studying climate change or historians examining California's agricultural legacy.
At The Bancroft Library, Christina Velazquez Fidler is addressing similar challenges around scale and access. As head of the library's Digital Collections Unit, she often works in the sprawling world of born-digital archives.
When a writer or artist donates their papers, the library may receive thousands of digital files, some containing intensely personal information.
Velazquez Fidler has used AI-generated Python scripts to search for sensitive data such as Social Security numbers. "The script lets us isolate those parts of the files, ... which for us has always been like finding a needle in a haystack," she said.
Staff members can then protect any sensitive items before the archive is made public. The process helps get materials into researchers' hands sooner—and, by reclaiming staff time, allows archivists to move more collections through the pipeline and into public view.
Protecting scholars' rights
While AI reshapes how collections are processed and discovered, it is also changing what scholars are able to do with information. Helping guide that work is the Library's Scholarly Communication and Information Policy team, led by Rachael Samberg.
One example of the team's impact centered on a simple question: Could scholars use AI to study how Hollywood films portray race and gender at scale? Until recently, they couldn't. Federal law prohibited breaking the digital encryption on DVDs, and streaming services barred downloading. There was no lawful way to build a dataset large enough for meaningful computational analysis.
Samberg and the SCIP team helped change that. Working with David Bamman, an associate professor in UC Berkeley's School of Information, the office submitted testimony to the U.S. Copyright Office, arguing for an exemption that would allow researchers to unlock DVDs for scholarly use.
In 2021, the exemption was granted. Bamman's team legally digitized more than 2,300 films and used AI to analyze 4,400 hours of footage—work that revealed a measurable increase in onscreen diversity since 2010.
The findings were recently published in an article co-authored by Samberg in Proceedings of the National Academy of Sciences.
The project captures SCIP's role in miniature: clearing legal barriers so scholars can use AI in ways that advance knowledge. But the office's work doesn't end with federal policy.
UC Berkeley licenses millions of dollars' worth of digital content each year, and if a publisher prohibits AI use or training, researchers are blocked from analyzing that content—even when fair use might otherwise allow it. SCIP negotiates for the rights scholars need, ensuring that contracts keep pace with researchers' interests.
"We are enabling research in foundational ways in an environment and landscape where the right to do things with information is under threat," Samberg said. "As a service, I think (our work is) both monumental and inspirational for what the Library can achieve to benefit Berkeley scholars."
The path forward
Even with the promise of new tools, library staff members are clear-eyed about the challenges ahead. AI raises tough questions—from environmental impact to copyright concerns—and the speed of adoption can feel dizzying.
"These reservations are understandable," Velazquez Fidler said. "And I think we ignore them at our own risk."
On her team, she is careful to use AI only for tasks that ease the workload for colleagues and often checks in with them about their comfort levels.
Across the Library, that same sense of responsibility is shaping how staff members think about the future.
Caizzi and a group of librarians and staff members are developing an AI framework to help the Library navigate this moment with intention—clarifying how the organization can use AI strategically, collaborate with campus partners, and support students as the landscape shifts.
To build that framework, Caizzi led a series of focus groups and listening sessions with Library staff. The goal was to surface concerns, gather use cases, and understand where AI might help—and where it might introduce new risks. Those conversations will inform a set of guiding principles around human agency, transparency, privacy, and equitable access, ensuring that any use of AI reflects the Library's values and the needs of its users.
The path forward is not without complexity. But many staff members are cautiously optimistic about the opportunities: the chance to improve accessibility for people with disabilities, to open collections in ways that were once unimaginable, and to accelerate research that could make the world better.
"AI is changing the landscape faster than any of us expected, but libraries have always adapted," Caizzi said. "Our job is to meet this moment with care, with clarity, and with the values that have guided us for generations."
Who's behind this story?
Swati Mestri holds a bachelor's degree in Electronics Engineering and has worked as a content editor since 2019. She has experience editing research documents across technology, health care, and materials science, and has a particular interest in technology and space. Full profile →
Bachelor's in mathematical biology, Master's in creative writing. Well-traveled with unique perspectives on science and language. Full profile →
Citation: Checking out AI: University library puts emerging technology to the test (2026, August 6) retrieved 6 August 2026 from https://phys.org/news/2026-08-ai-university-library-emerging-technology.html
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.
>> Home