I was scrolling through library news early this week, the way I do most mornings with my tea, and something struck me. Six or seven stories, from six or seven different corners of the world, all circling the same question without quite naming it: who is in charge of AI in a library, and what happens if nobody decides on purpose?
Let me walk you through what I found, because I think it matters more to us in India than the datelines suggest.
Start in Chicago. At the ALA Annual Conference 2026, a panel sat down to talk about the ethical use of AI in libraries, and the conversation went exactly where I expected it to go: privacy, transparency, intellectual freedom, equitable access. Nothing new there, you might say. Librarians have guarded these values for a century. But here is the twist. The panel’s real message wasn’t a list of principles, it was a single instruction: don’t adopt AI simply because it exists on a shelf. Ask what it does to your users first. I have said something close to this for two years now to anyone who will listen, so forgive me if I nodded a little too hard at my screen.
Meanwhile, Fairfield University quietly updated its AI Literacy Guide, and this one deserves more attention than it will get. The guide tells students to use AI for generating ideas, sharpening research questions, exploring topics they don’t yet understand. Fair enough. But it also insists, in plain language, that AI-generated content should never be accepted without review. That single line is, in my view, the whole future of reference service compressed into one sentence. We spent decades teaching students to evaluate a source. Now we are teaching them to evaluate an answer that has no source at all until you go looking for one. I have called this the Fluency Trap in my own writing: an AI response reads so smoothly that we mistake the writing for the truth. Fairfield just built that lesson into policy.
Behind all this front-of-house discussion, something quieter is happening in technical services, and it happens to be a subject close to my heart. More than 860 cataloguers across 77 countries are now using AI-assisted tools to process MARC21 records from photographs and PDFs. I have run this experiment myself, at a much smaller scale, with a Llama model on my own machine, cataloguing books photographed on a mobile phone. The accuracy was good, not perfect, somewhere around three-quarters at the field level in my own testing. What struck me both times, once in my study and once reading this week’s numbers, is that the tool doesn’t replace the cataloguer’s judgement. It just clears the underbrush so the judgement has somewhere to stand.
And this is where a new figure starts appearing in library conversations: the AI Librarian. Not a science fiction character, just a colleague who understands information science and AI well enough to sit between the two. Designing literacy programmes, checking tools for bias, supporting academic integrity, writing institutional policy. Whether the job title ever appears on an appointment letter in India is beside the point. The skills already do. I would go so far as to say I have been doing a version of this job since I retired from NIC, minus the letterhead.
The University of North Carolina Libraries updated its guidance on AI for evidence synthesis, running something called a Library AI Studio with workshops and consultations. The Conference of European National Librarians is collecting input from national libraries on text datasets and AI models for a forthcoming white paper. And the San José Public Library keeps expanding its AI Center for Civic and Social Good, teaching ordinary members of the public what these tools actually do under the bonnet. Different countries, different budgets, same instinct: don’t just use AI, teach people to see through it.
So where does India stand in all this? Closer than we usually give ourselves credit for. I see workshops on AI cataloguing and discovery systems happening at library schools across the country. I see young colleagues experimenting with digital repositories linked to platforms like e-Granthalaya. What I don’t yet see enough of is the governance layer, the part where an institution sits down and decides, on paper, what AI is and isn’t allowed to touch in its library. Ethical frameworks, staff training, human oversight, community access. These aren’t Chicago’s problems or Fairfield’s problems. They are library problems, and they arrived here the same week they arrived there, whether we noticed or not.
I keep coming back to a small line I wrote for myself a while ago, my Librarian’s 3-Step Verification Protocol: verify the source, disclose the tool, validate the output. It sounds almost too simple to matter. But then again, so did “wash your hands,” until it saved a great many lives. Some rules earn their keep precisely by being obvious enough to actually follow.
The question this week’s stories leave me with isn’t whether AI will change libraries. That ship, as they say, has long sailed. It’s whether we will be the ones steering it, or whether we’ll simply be told, one policy document at a time, where it has already gone.
References
Library Journal, “Ethical Use of AI in Libraries,” ALA Annual 2026 coverage: https://www.libraryjournal.com/story/ethical-use-of-ai-in-libraries-ala-annual-2026
Fairfield University, DiMenna-Nyselius Library, Artificial Intelligence Literacy Guide: https://librarybestbets.fairfield.edu/c.php?g=1480414&p=11031941
Scholaro, “The AI Librarian”: https://www.scholaro.com/db/News/ai-librarian-college-383
UNC Libraries, AI and Machine Learning guidance: https://guides.lib.unc.edu/AI-ML
CENL (Conference of European National Librarians), AI in Libraries Network Group: https://www.cenl.org/text-datasets-and-ai-models-input-needed-from-european-national-libraries/
San José Public Library, AI Center for Civic and Social Good: https://www.sjpl.org/artificial-intelligence/