Two essays hit my inbox in the last few days. One warned that AI concentrates power. The other argued libraries already have the skills to govern it. Together they made me wonder why so many libraries are still standing on the sidelines.
The first article, Tressie McMillan Cottom’s piece, “A Reading List for Our Age of A.I.” in the New York Times, will really strike a chord with the AI resistance minded. Cottom argues that AI isn’t a product: it’s politics. It’s part of the system where money, data, and power are consolidating in the hands of a few unelected executives. That’s certainly not hyperbole.
It’s an important reminder why libraries should not ‘opt out’ of AI literacy programs.
If you really believe that AI is politics, then whoever explains it to the public is doing vital work for democracy, whether they admit it or not. Without an authentic, public service voice in the conversation, messaging falls to the vendors, the influencers hyping the service, and consultants cashing in on chaos. None of them answer to the public. None of them take an oath to intellectual freedom, privacy, or equitable access.
Both priorities are long-standing, well-defined missions of the public library. That’s not branding; it’s the vital distinction between an institution and a sales funnel.
If Cottom is right that power is concentrating, the answer isn’t fewer trusted, disinterested guides in the room. Our communities need more voices advocating for the public interest. Libraries have exactly the kind of public trust this moment requires. Too many are choosing to spend it on “Avoid AI” workshops instead.
I know the objection, because I have made the point myself myself: engaging with these tools feeds the power machine Cottom is warning us about. Every library patron steered toward ChatGPT or Gemini is a data point handed to the companies concentrating that power.
Fair, but many of our patrons are already there. We’re not steering anyone at this point. Consider what the alternative actually buys. Opting out doesn’t slow the machine down; it just means nobody in the room is asking the hard questions on your patrons’ behalf.
My career in libraries was not spent helping people avoid technology. The job was always helping patrons master it.
Which is exactly why the second piece I read this weekend matters more than the first: “How Library Science Principles Power Community AI Literacy” from PublicLibrariesonline.org by Jose Ruiz-Vazquez, assistant manager at the Milwood Branch of the Austin Public Library.
Ruiz-Vazquez put it better than I can: “We do governance all the time. We just don’t call it that.”
Controlled vocabulary, provenance, bias evaluation, and information stewardship are not a side-quest to AI governance. We’re applying the same set of skills to a different challenge. Librarians have been examining the provenance and reliability of information for a long time. That experience doesn’t fade when the data is generated by a mathematical language model.
The key point here, and it’s the one that should end this debate: Ruiz-Vazquez, offers a direct institutional challenge: libraries are already positioned to be the “verification layer for democracy” but only if they claim the role.
What does “getting in the game” actually require? Every major technology transition has been figured out by people willing to experiment in public. Consider applying the skills our libraries already possess:
· Provenance: Where did this information come from?
· Authority: Who produced it, and who owns the system?
· Privacy: What happens to the data a patron enters?
· Evaluation: What are the tool’s limitations, biases, and failure modes?
· Access: Who can use it, and who is locked out due to cost, language or disability?
· Context: What does this output ignore?
· Human judgment: When should a person, not a model, make the decision?
A practical approach begins modestly:
· Create a small, public AI-tool registry. Evaluate the five tools patrons ask about most. Document privacy policies, costs, proper uses and limitations identified.
· Integrate AI literacy into existing programs. Help job seekers use AI without losing their own voice. Help patrons examine vital records without treating a chatbot as ‘the expert’. Help students understand research and develop critical thinking skills.
· Build standards together. Libraries should collaborate on evaluation frameworks, documentation practices, and preservation methods. We do not need every branch to reinvent the wheel, (or, in this case, the hallucinated zoning ordinance).
Solutions like this are starting to take shape. Frisco Public Library is building AI programming with community partners rather than a single subject workshops. That’s a launch point, not a finished model. Nobody’s built the whole thing. Could it be that the infrastructure doesn’t exist yet because too many libraries decided the safer, more virtuous lane was standing outside it?
The library’s distinctive contribution is not technical novelty. It is public-interest information stewardship. You don’t need a six-figure budget. You only need staff willing to try the tools, document where problems and shortfalls, and refuse to accept denial as a valid position.
Caution has a place. I’m less sure about “resistance”. Refusal for the wrong tool, at the right moment, might indeed be the correct response. But refusal as the whole strategy is outsourcing your professional judgment to whoever’s willing to stand up before you.
Libraries have spent decades helping people navigate every major shift in how information is created, distributed, and trusted. AI isn’t the first revolution we’ve seen. It’s simply the first one that too many librarians have decided someone else should lead.
If we think our role in a public library is to teach people to avoid technology, we’ve misunderstood the assignment.