āĻĒāϰāĻŦāϰā§āϤ⧠āĻāĻĒāĻĒāϰāĻŦāϰā§āϤ⧠āĻāĻžāĻāĻĄ
āĻāĻāĻ āĻāĻŦāĻ āĻā§āĻĒāύā§āϝāĻŧāϤāĻž
āϏāĻŽāĻžāĻ
āϏā§āϏāĻžāĻāĻāĻŋ āĻāĻžāĻāĻĄ
AI features in library catalogs, discovery tools, chat services, or vendor platforms can process searches, reading histories, reference questions, and personal information.
āĻāĻ āύāĻŋāϰā§āĻĻā§āĻļāĻŋāĻāĻž āĻļā§āύā§āύāĻāĻāĻ-āĻŦāϰā§āĻŖāĻŋāϤ ¡ ~3 min
Libraries should understand and limit that data use, disclose AI-enabled services, and preserve a meaningful non-AI path when possible; protecting patron confidentiality remains central to library service.
Libraries have long treated reading, borrowing, searching, and reference questions as sensitive. AI can create new ways for those records to be stored, combined, inferred from, or shared. A library assistant may receive a question that reveals health, religion, political interests, immigration concerns, or a personal crisis. If a third-party tool retains prompts or uses them to improve a model, a patronâs private inquiry could leave the institutionâs control. Data may also flow through a catalog vendor, consortium, school district, campus, or municipality rather than a library-operated system. The American Library Associationâs 2026 Guidance on the Use of Artificial Intelligence in Libraries recommends reviewing what AI-enabled systems collect, whether features are optional, whether information is used for training, where it is stored, how long it is kept, and how it can be deleted. The guidance says libraries should disclose when a third party processes patron data or influences search and recommendations, explain choices, and provide a meaningful non-AI or minimally automated path when possible. It also advises against entering names, borrowing histories, reference interactions, or other non-public records into unapproved tools. Privacy review should occur before adoption and after meaningful vendor or configuration changes. Ask whether a feature is enabled by default, whether staff can disable it, which subprocessors receive data, and whether searches can be tied to an account. Contract terms should address training use, retention, deletion, access controls, breach notice, audits, and exit. A promise that data are âsecureâ is not a full answer to how they are used. Staff need plain guidance that protects patrons without preventing useful work. Personalization also shapes discovery. A recommendation system may narrow what a patron sees or infer interests from borrowing behavior. Libraries should retain predictable access paths such as known-item search, chronological lists, and subject browsing, and review results for missing viewpoints or languages.
A librarian declines to paste a patronâs named reading history into a public chatbot while helping them find new materials.
A library evaluates whether its discovery vendor uses search queries or borrowing records to train or personalize AI features.
A catalog offers AI-generated recommendations alongside known-item search and subject browsing that do not require a patron profile.
A library explains in plain language when AI processes queries and how patrons can turn off personalization or request staff assistance.
Library AI use will expand through embedded vendor features as well as tools purchased directly by institutions. Data flows can be hard to see when multiple systems share identity and search information. ALA guidance and privacy practices will evolve as interfaces change. Libraries should reassess vendors, configurations, and retention on a schedule and after major updates. Trust depends on clear notice, genuine choice, minimized data, and preserving staff assistance and open discovery. Privacy controls shape whether patrons can safely explore sensitive questions. If people cannot tell who can see a search or borrowing record, they may avoid using that service; explain the data boundary before they enter a query.
āĻĢā§āϰāĻŋ āύāĻŋāĻāĻāϞā§āĻāĻžāϰ
āĻĒā§āϰāϤāĻŋ āϏāĻĒā§āϤāĻžāĻšā§āϰ āĻāĻžāĻā§āϰ āĻĻāĻŋāύ āϏāĻāĻžāϞ⧠āϤāĻŋāύāĻāĻŋ āϝāĻžāĻāĻžāĻāĻā§āϤ āĻāĻāĻ āϏā§āĻā§āϰāĻŋ, āϏāĻšāĻ āĻāĻāϰā§āĻāĻŋāϤ⧠āϞā§āĻāĻžāĨ¤ āĻāĻŋāϰāĻāĻžāϞā§āϰ āĻāύā§āϝ āĻĢā§āϰāĻŋ, āĻā§āύ⧠āĻŦāĻŋāĻā§āĻāĻžāĻĒāύ āύā§āĻāĨ¤
āĻĒā§āϰāϤāĻŋ āϏāĻĒā§āϤāĻžāĻšā§āϰ āĻāĻžāĻā§āϰ āĻĻāĻŋāύ⧠āĻāĻāĻāĻŋ āĻāĻŽā§āϞāĨ¤ āĻāĻ āĻā§āϞāĻŋāĻā§āĻ āĻāύāϏāĻžāĻŦāϏā§āĻā§āϰāĻžāĻāĻŦ āĻāϰā§āύāĨ¤ āĻāĻŽāϰāĻž āĻāĻāύā§āĻ āĻāĻĒāύāĻžāϰ āĻ āĻŋāĻāĻžāύāĻž āĻŦāĻŋāĻā§āϰāĻŋ āĻŦāĻž āĻļā§ā§āĻžāϰ āĻāϰāĻŋ āύāĻžāĨ¤
āύāĻŋāĻā§āĻā§ āĻĒāϰā§āĻā§āώāĻž āĻāϰā§āύ
āĻĒā§āϰāϤāĻŋāĻāĻŋ āĻāϤā§āϤāϰā§āϰ āϤāĻžā§āĻā§āώāĻŖāĻŋāĻ āĻĢāĻŋāĻĄāĻŦā§āϝāĻžāĻ, āĻāĻŦāĻ āĻā§āϰā§āϏ āĻĒāĻžāϏ āĻāϰāĻžāϰ āĻĒāϰ āϝāĻžāĻāĻžāĻāϝā§āĻā§āϝ āĻāĻāĻĄāĻŋ āϏāĻš āĻāĻāĻāĻŋ āĻļā§ā§āĻžāϰāϝā§āĻā§āϝ āϏāĻžāϰā§āĻāĻŋāĻĢāĻŋāĻā§āĻāĨ¤
āĻĢā§āϰāĻŋ āĻāĻāĻ āĻļāĻŋāĻā§āώāĻžāĻā§ āϏāĻŽāϰā§āĻĨāύ āĻāϰā§āύāĨ¤ AI Understanding āĻāĻāĻāĻŋ 501(c)(3) āĻ āϞāĻžāĻāĻāύāĻ āĻĒā§āϰāϤāĻŋāώā§āĻ āĻžāύ â āĻā§āύ⧠āĻŦāĻŋāĻā§āĻāĻžāĻĒāύ āύā§āĻ, āĻā§āύ⧠āĻĒā§āĻā§āĻžāϞ āύā§āĻ, āĻāĻāύā§āĻ āύāĻžāĨ¤ āĻāĻāĻāĻŋ āĻĻāĻžāύ āĻāϰā§āύ
AI features in library catalogs, discovery tools, chat services, or vendor platforms can process searches, reading histories, reference questions, and personal information. Libraries should understand and limit that data use, disclose AI-enabled services, and preserve a meaningful non-AI path when possible; protecting patron confidentiality remains central to library service.
Borrowing history is sensitive and should not be entered into an unapproved system.
ALA guidance emphasizes clear disclosure, choices, and alternatives where possible.
Training terms affect where patron information may flow and how it is reused.
Non-personalized paths let people discover materials without a profile.
Other data fields can be linkable or reveal sensitive behavior.
āĻļāĻŋāĻāϤ⧠āĻĨāĻžāĻā§āύ
āĻāĻ āĻŦāĻŋāώāϝāĻŧā§āϰ āĻāύā§āϝ āĻŦāĻžāĻāĻžāĻ āĻāϰāĻž āĻāϰāĻ āĻāĻžāĻāĻĄ
āĻĒāϰāĻŦāϰā§āϤ⧠āĻāĻĒāĻĒāϰāĻŦāϰā§āϤ⧠āĻāĻžāĻāĻĄ
āĻāĻāĻ āĻāĻŦāĻ āĻā§āĻĒāύā§āϝāĻŧāϤāĻž
āϏāĻŽāĻžāĻ