AI Governance in GLAM / Infophilia Tools

Infophilia Tools · Companion to Core Readings · iSchool Illinois

AI Governance Across Libraries, Archives & Museums

Reading professional AI statements as governance documents — and as information-organization documents.

A comparative reading of selected professional statements, guidance, and publication policies on artificial intelligence, machine learning, and generative AI. Its purpose is not to restate every policy, but to help readers compare how each profession frames adoption, refusal, accountability, and human authority. A short set of contextual comparators then situates the corpus — the international bodies ICA, ICOM, and UNESCO, and, from outside the memory professions entirely, the Authors Guild.

This is a selective, annotated comparison — not a comprehensive inventory, and not legal, procurement, or institutional-policy advice. It distinguishes among (1) AI and machine learning in services and metadata workflows, (2) generative AI in teaching, research, and publication, and (3) governance questions involving privacy, labor, intellectual freedom, copyright, accessibility, environmental effects, and public trust.

Method & currency. Each entry is based on the linked official document or organizational page. “Position on use,” “safeguards / refusal,” and “distinctive contribution” summarize each document’s stated approach; they are A. S. Coleman’s comparative interpretation, not official organizational language. Reviewed August 2026. Because AI guidance changes quickly — ALA’s guidance, for one, is an explicitly living document — check the linked source for the current version.
Position on use (adoption)
The document identifies legitimate or beneficial uses of AI.
Safeguards / refusal
The document names conditions under which use should be limited, rejected, disclosed, or subject to human review and accountability. A document can support limited AI use while refusing to automate judgment, authorship, interpretation, or responsibility.
Governance object
What the document actually governs — the thing AI is being brought to bear on. This is where the comparison becomes interesting: the documents are not all governing the same thing.
1

Sector-wide library guidance

Professional associations speaking to the field: how libraries should decide, what rights they retain, and what social role they hold.

Library-sector guidance. Compiled and interpreted by A. S. Coleman; positions summarized, not quoted.
Organization & documentGovernance objectPosition on useSafeguards / refusalDistinctive contribution
ALAGuidance on the Use of Artificial Intelligence in Libraries (2026). U.S. libraries; local policy, procurement, vendor integrations.Institutional AI adoption in services & operationsConditional, purpose-driven use when a documented public-service purpose exists and benefits and risks have been assessed.Don’t adopt AI merely because it is available; evaluation outcomes explicitly include adoption, limited use, pilot, delay, pause, non-adoption, removal, or discontinuation; retain human judgment and accountability; assess labor, privacy, accessibility, sustainability, legal, vendor, and trust effects. A living document adopted by ALA Council in July 2026.Most explicit adoption-to-refusal decision framework in the U.S. library sector; grounds AI decisions in public good, intellectual freedom, privacy, sustainability, DEIA, and labor.
ARLResearch Libraries Guiding Principles for Artificial Intelligence (April 2024). Research libraries; licensing, literacy, scholarly environment.Information rights, licensing & scholarly infrastructureAI can expand access, openness, and AI literacy across the research environment.“No human, no AI” — human involvement at critical decision points; advocate transparency in algorithms, training data, and methods; protect privacy and security; address bias; negotiate against licenses that restrict scholarly use.Strongest treatment of libraries as rights advocates, license negotiators, and stewards of scholarly access.
IFLAStatement on Libraries and Artificial Intelligence (2020). International library-sector statement.Professional & societal role of librariesSupports libraries as active, critical participants in the development and use of AI.Calls for ethical oversight, AI literacy, and professional capacity, and for safeguards before delegating resource-description or access functions to AI; emphasizes intellectual freedom and privacy.One of the earliest major international library statements; foregrounds social role and professional responsibility over vendor adoption.
2

Archives & publication-integrity policies

Not statements about archival practice in general, but governance of a publishing program — where the threshold rises as AI moves from assistance toward authorship.

SAA publication-integrity policies. Positions summarized, not quoted.
Organization & documentGovernance objectPosition on useSafeguards / refusalDistinctive contribution
SAA — American ArchivistGenerative AI Statement, Editorial Board (2024). Journal editorial policy for articles and reviews.Scholarly journal publicationAI may be used; use alone does not trigger rejection.Purely AI-generated content is prohibited; AI use must be disclosed; editors may request prompts or transcripts; human review remains decisive.A concrete publication-integrity model: disclosure, traceability, and human authorship.
SAA — Publications BoardGenerative AI Statement (2024). SAA book and publication submissions.Book publication & authorshipAI may assist under transparent conditions.Proposals, drafts, and final publications must be conceptualized and written by human beings; disclosure required; editors may request transcripts.A higher human-authorship threshold than the journal statement — read together, the pair shows governance tightening as AI moves from assistance toward authorship.
3

Field-level statement (information science)

Not a procurement manual or a conduct code, but a statement about the field’s own research ethics — who gets to produce knowledge about AI, and whose harms and perspectives are represented.

Field-level statement. Positions summarized, not quoted.
Organization & documentGovernance objectPosition on useSafeguards / refusalDistinctive contribution
ASIS&T, ALISE & iSchoolsStatement on AI ethics and the contributions of diverse voices (2020). Joint field-level statement for information-science research and discourse.AI-ethics research & whose voices are representedSupports rigorous research to enable ethical AI applications.Insists that diverse voices and critical research are essential, especially where AI produces discriminatory or harmful outcomes; attends to algorithmic flaws, bias, and transparency in dissemination.A field-level statement on AI ethics, research, and diversity of perspectives, issued amid the controversy surrounding Dr. Timnit Gebru and debates over AI research, bias, and scholarly voice — not an implementation or procurement framework.
4

Museum-sector guidance & emerging practice

Mission fit, public disclosure, and labor — with guidance still actively developing on both sides of the Atlantic.

Museum-sector guidance. Positions summarized, not quoted.
Organization & documentGovernance objectPosition on useSafeguards / refusalDistinctive contribution
Museums Association (UK)An ethical approach to AI (2024); AI provisions in the Code of Ethics (2025); AI & Ethics Expert Working Group (2026). Museum-sector practice guidance.Museum practice, mission & laborUse AI only where it aligns with the museum’s mission and existing policies.Publish a public AI-use statement; evaluate ethics before use; weigh effects on employment in an under-resourced sector; attend to bias and decolonial concerns.Foregrounds mission alignment, public disclosure, and labor. The 2024 guide is not the endpoint: a 2026 working group is developing further guidance.
Canadian Museums AssociationEmerging AI ethics direction. Sector discussion / developing guidance.Sector ethics / emerging policyAcknowledges the potential of AI in museum work.Points toward informed consent, privacy, transparency, bias testing, and revising ethics guidance to include AI.Centers visitors, cultural trust, data consent, and representational concerns — still developing rather than a finished framework.
5

Institutional implementation example

A national institution, not a professional association — useful because it shows governance operationalized against real collections and workflows.

Institutional implementation. Positions summarized, not quoted.
Organization & documentGovernance objectPosition on useSafeguards / refusalDistinctive contribution
Library and Archives CanadaAI approach under Vision 2030 (Departmental Plans 2025–26, 2026–27). National GLAM institution; implementation-oriented.Collections, description, discovery & operationsAI aids transcription and handwriting recognition (e.g., Transkribus), metadata optimization, discovery, and other time-consuming workflows.Human review for access and interpretation; privacy, copyright, and collection integrity; deliberate attention to accurate, respectful description; developing its AI approach within ongoing institutional planning.One of the clearest operational GLAM examples, connecting AI directly to metadata, description, discovery, collections, professional expertise, and public trust.
§

Contextual comparators beyond the GLAM corpus

These bodies are not part of the nine-row comparison — they do different work at different scales — but they situate it. Two are the international professional umbrellas standing above the documents above; one is a cross-sector normative instrument; one is a rights-holder comparator from outside the memory professions entirely.

ICA — International Council on Archives
International archival professional context. The global body for archives and archivists (founded 1948, Paris); its WIPO position statement on intellectual property and AI stresses access to information, transparency, and archival exceptions, and urges conversations beyond rights-holders alone. ica.org →
ICOM — International Council of Museums
International museum professional context. The global museum body; its revised Code of Ethics (2026) addresses contemporary challenges, including AI, through an ethical lens. icom.museum →
UNESCO — cross-sector normative context
Its Recommendation on the Ethics of Artificial Intelligence (2021) is the first global normative instrument on AI ethics, adopted by all member states, and a reference point beneath much of the sector guidance above. unesco.org →
Authors Guild — human-authorship & intellectual-labor comparator
Represents the authors whose works become training data, arguing for consent, compensation, and credit, and that mass copying to train AI is not fair use. Parallel positions come from the UK Society of Authors and the European Writers’ Council. authorsguild.org →

What the comparison reveals

GLAM does not have one AI governance problem; it has different governance objects. Libraries ask whether to adopt a system and under what conditions; research libraries ask what rights they retain when AI enters the scholarly record; archives ask what counts as human authorship; information science asks whose harms and perspectives are represented; national institutions ask what happens when AI touches description and access; museums ask whether AI fits the mission and what it does to labor.

One comparator set stands apart from all of these: authors’ and writers’ organizations — the Authors Guild, the Society of Authors, the European Writers’ Council — govern the inputs, asking whether works may be used to train AI at all, and on what terms of consent, transparency, and compensation. Set beside the institutions that deploy AI, they surface the question the memory professions cannot answer alone: who consented to the training data?

The sharper divide is not pro-AI versus anti-AI — almost none of these documents are simply against AI. It is what each profession treats as non-delegable to AI: judgment, authorship, interpretation, description, or responsibility.

❦

Questions for readers

  • What work should remain human-led because it requires judgment, accountability, contextual knowledge, or authorship?
  • What evidence should an institution require before adding AI to discovery, metadata, reference, instruction, or collections workflows?
  • Who can inspect, contest, or correct the systems that shape knowledge access?
  • How should libraries, archives, and museums assess effects on labor, privacy, accessibility, sustainability, cultural stewardship, and public trust?
§

Citation & reuse

Suggested citation: Coleman, A. S. (2026). AI governance across libraries, archives, and museums: A comparative reading (Version 0.1). Infophilia Tools. https://infophilia.codeberg.page/infophilia-tools/ai-governance-glam/

Original annotations and comparative analysis are licensed CC BY-NC-SA 4.0. Linked documents remain subject to their own terms. This is a companion to Core Readings for Information Organization & Access →