The Way I See It

Anna Małgorzata Kamińska

When the Program Becomes the Author

A Librarian’s View from the Threshold

Anna Małgorzata Kamińska is an academic librarian at Silesian University of Technology (Gliwice, Poland) and an assistant professor at the University of Silesia (Katowice, Poland), email: anna.kaminska@us.edu.pl.

© 2026 Anna Małgorzata Kamińska

I watched Tron: Ares not long ago, a film about a digital program that was engineered to comply—and chose instead to author. Ares crosses the boundary between its computational world and ours and in doing so becomes something its creators did not design: not a tool but an agent with its own logic, its own purposes, and its own claim on what is real. I found myself thinking less about the film and more about my library.

I hold what may seem like contradictory positions. I am an academic librarian at Silesian University of Technology in Gliwice, Poland, where I am currently implementing generative artificial intelligence (AI) tools into our library services. I am also an assistant professor in library and information science at the University of Silesia. I am, in other words, both installing the infrastructure and studying what it does to us—which turns out to be an unsettling place to stand.

The AI tools I am implementing are impressive. They synthesize literature. They answer complex research questions. They respond in the user’s language with the user’s framing. And therein lies the shift I have been struggling to articulate.

When students ask a generative AI system to summarize the state of the field rather than searching our databases, they are not bypassing a search tool. They are bypassing an epistemic institution—one that has historically organized the conditions under which knowledge claims are validated, evaluated, and made trustworthy. The difference matters more than we have been willing to say.

Three things are happening simultaneously. First, disintermediation: Users now receive synthesized, interpreted knowledge directly from AI systems, without library mediation. Second, infrastructure dependency: The training data on which these AI systems operate was assembled outside any institution with a public mandate for knowledge quality—including ours. Third, authority transfer: The epistemic trust that students and researchers once extended to curated, accountable sources is migrating to algorithmic outputs whose provenance is opaque. This is not a technological problem. It is an epistemological one.

When a peer reviewer recently responded to an article my colleagues and I had submitted on generative AI in scholarly communication, the reviewer’s argument stopped me—not because it dismissed our concerns but because it forced me to recognize how far the transformation had already proceeded. The reviewer treated the library’s predicament not as a crisis to be averted but as a condition to be theorized. The crisis, the argument implied, had already happened. We were writing its analysis, not its prevention.

That moment, combined with the daily experience of implementing AI tools in a library I know deeply, crystallized something I had been circling for months.

Academic libraries are not facing a technological transition. They are facing a constitutional moment—a point at which an institution must decide what it actually is. The choice is not between embracing AI and resisting it. It is between defining your institutional identity in terms adequate to what AI has become—an epistemic actor, not a search engine—or accepting a residual role while the infrastructure of knowledge generation is constituted around you.

Epistemic marginalization is not the structural destiny of academic libraries in the age of generative AI. It is the outcome of failing to act as though it is not.

What does acting look like? I see three places where libraries already hold leverage and have not yet fully claimed it.

The first is data stewardship. The institutional repositories we manage are among the most carefully curated collections of research outputs in existence. As AI training corpora become the substrate of knowledge generation, the question of who governs what enters them—and on what terms—becomes urgent. Libraries that actively shape their data infrastructure are already participants in AI’s epistemic future, whether or not they recognize themselves as such.

The second is governance. Every university is currently writing its AI policy. Most of those conversations are happening in IT committees and administrative offices, with legal counsel and risk management. Librarians—with decades of practice in source evaluation, information ethics, and the principles of fairness, accountability, transparency, and ethics—are largely absent from these tables. This is not an oversight we should continue to accept.

The third is critical AI literacy. Not teaching users to operate AI tools but teaching them to evaluate what AI produces: whose assumptions it encodes, what it systematically excludes, when its confident synthesis conceals an absence of evidence. This is information literacy. It is, in the most direct sense, what we do—and it is what no AI system can provide about itself.

In Tron: Ares, the digital entity does not defeat the system by resisting it. Having crossed the threshold, having become something more than its creators intended, Ares must negotiate the terms of its existence in the world it now inhabits. The film’s resolution is not suppression but governance: the question of what kind of actor AI becomes and in whose institutional framework.

The question before academic libraries is not whether the program becomes the author. It has. The question is whether the library becomes its co-architect or consents, by inaction, to become its archive.