In the previous post in this series, I explored what happens when knowledge systems begin to act on what they know. As AI moves further into discovery platforms, repositories, catalogs, and workplace systems, another question follows: how should organizations respond when a technically capable system conflicts with the principles they exist to uphold?
That question has acquired fresh urgency. In June 2026, the American Library Association (ALA) published its Guidance on the Use of Artificial Intelligence in Libraries. The document offers a values-based approach to selecting and using AI tools, covering systems embedded in databases, catalogs, productivity tools, and other vendor platforms.
The guidance is aimed at libraries, but the underlying question extends across archives, museums, and knowledge management more broadly. These organizations hold different collections and serve different communities. However, they share responsibilities for protecting information, preserving context, supporting access, and exercising professional judgment.
So, what happens when technology begins to challenge those responsibilities?
What Values Are Being Challenged?
Professional values can sound abstract until technology forces an organization to reconsider its choices. The ALA guidance identifies public good, intellectual freedom, privacy, sustainability, diversity, equity, inclusion, and access, and labor as considerations in AI decisions.
The Society of American Archivists emphasizes access, accountability, preservation, responsible custody, and reliable records, while the International Council of Museums describes museum collections as a public trust requiring documentation, accessibility, provenance, and responsible stewardship.
Although the language varies, these institutions share commitments to access, privacy, collection integrity, and human accountability.
Public expectations reflect some of those differences in emphasis. An analysis reported by Heritage Social found greater enthusiasm for creative uses of AI in museums and galleries, more ambivalence about its effect on librarianship, and stronger concern about authenticity in archives.
The findings come from social media discourse rather than a representative survey, but they illustrate how responses depend on what people expect information and cultural institutions to protect. An application that appears imaginative in a gallery may cause concern when applied to an archival record.
AI places these professional responsibilities—and the public trust attached to them—under pressure.
How Values Affect Technology Decisions
The ALA guidance is significant because it connects professional principles with practical choices. It asks libraries to consider an AI application’s purpose, its effects on users and workers, its environmental impact, the safeguards available, and whether a non-AI option should remain available.
Transparent Access
Access is a good example. An AI-powered discovery tool may help users navigate a large collection or overcome language and accessibility barriers. However, automated ranking and personalization also influence which materials become visible.
If users cannot understand how results were selected, they cannot assess how the system may be shaping their access to information. Where possible, institutions should provide clear disclosure, retain predictable non-personalized routes through collections, and allow users to opt out of AI assistance.
Personalized Services and Privacy
Personalized services depend on processing information about their users. On streaming and shopping platforms, many people accept this as the price of tailored recommendations and greater convenience, although they may have little understanding of how extensively their data is collected or reused.
For libraries and knowledge services, the implications are more sensitive. Membership records, borrowing and reservation histories, search activity, research requests, support tickets, and reference inquiries can reveal far more than basic account details. They may indicate a person’s professional interests, health concerns, political views, or private circumstances.
Public services have a particular responsibility to scrutinize AI-enabled services. They need to understand what data vendors collect, where it is stored, how long it is retained, and whether searches, prompts, or account activity may be used to train or improve external models.
Context and Accuracy
Stewardship is challenged when AI-generated descriptions, summaries, or classifications enter institutional systems. Automated tools can increase capacity, but fluent output does not guarantee reliable provenance or appropriate context.
Much of the research behind this series emphasizes the need for mediated workflows, in which AI prepares suggestions while qualified professionals retain authority over the final record. Human judgment is essential, but it must be meaningful: if staff can review an output but lack the authority to correct, reject, or escalate it, human oversight exists in name only.
What Happens When Institutions Draw a Line?
One of the most significant features of the ALA guidance is its recognition that removing or declining an AI tool does not represent a failure of adoption. A considered decision to delay, limit, or discontinue its use can be an affirmative professional choice grounded in library values.
This matters because refusal to adopt a tool is often portrayed as resistance to innovation. Yet organizations may be right to decline technologies that conflict with their legal duties, service standards, or policies. Pressure may also come from outside an institution’s own systems. AI-generated content can affect national collections, resources, and established responsibilities.
For example, in May 2026, the Korean National Assembly amended the Library Act to allow the National Library of Korea to decline certain AI-generated publications after formal review. The change responded to rapidly produced “one-click publications” used to claim legal-deposit payments.
For present purposes, this case shows how generative AI can expose assumptions within an established national process. It also raises larger questions about preservation and cultural value, which I will return to in a future article on AI and national legal deposit.
Drawing a line can mean limiting automation, retaining human approval, restricting the use of sensitive data, declining a system altogether, or even raising questions in national parliaments. The appropriate boundary depends on the institution’s responsibilities and the consequences if its safeguards fail.
Wider Relevance Across GLAM and Knowledge Management
The ALA guidance provides a timely library-related example, but similar choices are emerging across the wider GLAM and KM world.
Archives must consider whether automated description preserves the context and reliability of records. Museums need to decide how AI-generated interpretation affects cultural authority and the communities represented in their collections. Knowledge teams must determine which repositories an AI system can access and which actions remain subject to approval.
Across these settings, the same practical options recur. Institutions may need to limit automation, preserve human judgment, protect information and user data, evaluate technology against professional principles, and refuse applications that cannot be reconciled with those principles.
These decisions require people who understand both the technology and the institution. Technical performance alone cannot determine whether a system supports intellectual freedom, respects the meaning of a collection, or serves the public interest.
A Conversation That Is Only Beginning
Institutional values cannot resolve every technology decision, but they reveal what is at stake and what an organization is responsible for protecting.
The ALA guidance is valuable because it turns established principles into practical criteria for procurement, system design, oversight, and non-adoption. Its wider challenge applies across the information and GLAM sectors: institutions need to define where AI supports their purpose, where safeguards are required, and where a boundary should be drawn.
This article opens that conversation. Automated discovery, open access, sustainability, and the impact of AI content on national collections all warrant closer examination.
Putting these principles into practice also depends on the people responsible for interpreting them. In the next article in this series, we will turn to the future of information work and consider how AI is changing professional responsibilities, the value of human expertise, and the authority information professionals need to shape its use.
0 Comments