AI Governance Has a Library Problem
“AI governance” has become a catch-all phrase in D.C., attached to a growing stack of bipartisan bills. But for all the activity, the conversation is oddly limited. Governance is usually understood as legislation, agency rulemaking, or corporate compliance programs. One group of decision-makers is almost never mentioned: librarians.
This omission is striking. AI systems are built on information infrastructure: data, classification systems, retrieval mechanisms, and access controls–precisely the domains libraries have governed for decades. Librarians have long served as intermediaries between information systems and the public, balancing access, privacy, accuracy, and harm. As new technologies emerge, librarians are often on the front lines of integration and interpretation.
Today, they are making key decisions about how generative AI tools appear in classrooms, public spaces, and research environments. In practice, they are governing AI. If AI governance is to be effective, the expertise of librarians–especially those working within legal and public institutions–must be recognized as central to how AI is governed.
Libraries Govern Information, and AI Is Information Infrastructure
Libraries have long served as champions of equitable access to information, privacy, classification and discoverability, and responsible approaches to harmful or misleading content.
As AI has gained mainstream popularity, librarians have continued this work, adapting longstanding governance practices to new technologies.
Libraries have operated electronic research assistants and automated systems for years, using them to manage circulation or support research while maintaining human oversight and institutional safeguards.
Far from being passive adopters, librarians have also taken the lead in shaping norms around AI use through formal efforts.
At the national and regional levels, librarians, including law librarians, have convened and created LibGuides and white papers to grapple explicitly with the implications of AI for legal research, privacy, and professional judgment.
Initiatives such as the Artificial Intelligence and the Future of Law Libraries South-Central Roundtable and the Future of Law Libraries convenings reflect a sustained effort by librarians to assess risk, set boundaries, and articulate best practices.
Why Policymakers Overlook Libraries
Despite the depth of expertise that libraries consistently demonstrate, policymakers rarely offer them a meaningful seat at the table in AI governance discussions. Law librarians–professionals trained at the intersection of legal analysis and information stewardship–are similarly sidelined.
Instead, industry experts and national or corporate governance frameworks are treated as the primary sources of authority, while governance that occurs at the library level is overlooked or dismissed.
This marginalization may reflect a persistent misperception: librarians are equipped to govern libraries, but not information itself.
Yet information (how it is classified, accessed, interpreted) is precisely what AI systems produce and mediate. Undervaluing librarians misunderstands both the nature of AI and the longstanding role libraries play in managing information.
Policy Lessons from Law Libraries
Policymakers can learn from librarians broadly, but law librarians are uniquely positioned to inform AI governance. Law librarians are professional interpreters of legal complexity, charged with weighing authority and making sense of dense, interconnected legal information systems.
If we want smarter AI governance, we should start by looking at what law librarians are already doing. From the American Association of Law Libraries’ legislative priorities to practical research guides, their experience offers clear lessons for governing AI.
- Human mediation matters. AI output is not authoritative and must be interpreted within legal and institutional context.
- Refusal is governance. Choosing not to deploy an AI system is a legitimate, and often responsible, regulatory outcome.
- Legal context cannot be automated. Especially in domains affecting rights, benefits, or liability, judgment cannot be delegated to machines.
Recognizing these lessons would move AI governance beyond abstract compliance frameworks and toward institutions that already manage information responsibly.
Ultimately, AI governance does not need to be invented; it needs to recognize and elevate the institutions already doing it well.
Author Bio: Nina-Simone Edwards, J.D. is a Senior Institute Associate at Georgetown University Law Center’s Institute for Technology Law & Policy and a 2025 #ShareTheMicInCyber Fellow at New America. Before joining the Tech Institute, Nina-Simone was a Fulbright Scholar and a City Year AmeriCorps Member. She will soon graduate with a Master’s in Library and Information Sciences and will become a law librarian.
