1 of 18

Thinking Through AI Policy for your Library

2 of 18

Why AI Policy Matters

auto_awesome

AI is already here

Generative AI tools are already being used by patrons and staff. A policy provides the necessary framework for current usage.

straighten

Consistency matters

Guidelines ensure a uniform approach across different departments, preventing fragmented or contradictory practices.

verified_user

Protect values

A robust policy safeguards core library values such as privacy, intellectual freedom, and equitable access in the digital age.

3 of 18

This Template Is a Starting Point

settings_suggest

Flexible

Easily customize sections to meet your library's unique operational needs.

published_with_changes

Adaptable

Designed to evolve alongside rapidly changing AI technologies and capabilities.

gavel

Not legal code

Consult with legal counsel to ensure compliance with local laws and regulations.

https://www.rhastings.net/ai-resources/ - Halfway down the page, under Templates From Robin

4 of 18

Purpose & Scope

flag

Purpose

Defines the foundational reasoning and intent—establishing clearly why the policy exists and what it aims to achieve within the organization.

groups

Scope

Outlines the boundaries of the policy, specifying exactly who it applies to and the specific circumstances or departments covered.

Most libraries won’t need to change much here—just insert your name and decide who is covered. You might, however, want to add a definition of AI (or GenAI) here so that everyone knows what the terms you are using actually refer to. I’ve got one from the Oxford English Dictionary in the long version you are welcome to grab, or you can come up with your own that works for your library. If you’d prefer to include more definitions than just the general AI one, there is a section (12) at the end where you can add definitions of things like what constitutes confidential information and what the differences between AI and GenAI are, along with any other jargon-y terms that might end up in this document.

1. Purpose

The purpose of this policy is to provide guidance for the ethical, responsible, and effective use of Artificial Intelligence (AI) within [Library Name]. Our definition of AI is from the Oxford English Dictionary: “software used to perform tasks or produce output previously thought to require human intelligence, esp. by using machine learning to extrapolate from large collections of data” (Oxford English Dictionary)

This policy ensures that AI:

Enhances—not replaces—human expertise

Protects patron privacy and intellectual freedom

Supports equitable access to information and services

Is used transparently and responsibly

AI is a tool. Final responsibility for decisions, services, and content remains with library staff.

2. Scope

This policy applies to:

All library staff, volunteers, and contractors

All AI tools used in library operations or services

All patron-facing and internal uses of AI

[OPTION: Include board members if relevant]

5 of 18

Guiding Principles

person

Human-centered

lock

Privacy

visibility

Transparency

assignment_turned_in

Accountability

If staff remember one section, it should be this. It defines your values. The accountability principle states that staff take responsibility for the output of AI that they use - it puts humans as responsible agents, since AI can’t do that (yet?) The H-A-H workflow is included here, you can keep or remove as you’d like. Other principles that you can adjust include Equity and bias awareness - feel free to add your own

3.1 Human-Centered Approach

AI will be used to support human judgment, not replace it. Human oversight is required in all meaningful decisions.

3.2 Human–AI–Human Workflow

Library use of AI will follow this model:

Human Input

Staff define the task, context, and goals

AI Assistance

AI generates suggestions, drafts, or analysis

Human Review and Decision

Staff verify, edit, and take responsibility for final output

AI outputs must always be reviewed before use.

3.3 Privacy and Confidentiality

No confidential or personally identifiable patron data may be entered into non-approved AI systems

Staff must follow all existing privacy and confidentiality policies

If in doubt: do not input the data

Protecting patron privacy is paramount.

3.4 Transparency

The library will disclose when AI is used in services or content

Patrons should be informed when interacting with AI systems

3.5 Equity and Bias Awareness

AI tools will be evaluated for bias and fairness

Staff will actively mitigate harmful or discriminatory outcomes

3.6 Accountability

Staff are responsible for AI-assisted work products

AI does not replace professional judgment or ethical responsibility

6 of 18

Human–AI–Human Workflow

person

Human defines

The person sets the objective, direction, and parameters of the task.

psychology

AI assists

Artificial Intelligence processes information and generates a draft or solution.

fact_check

Human reviews

The human verifies, edits, and finalizes the AI-generated output for quality.

7 of 18

Privacy

lock

No confidential data

Protect sensitive information by ensuring that no restricted or confidential materials are shared within the tool.

help_outline

When in doubt, don’t input

Adopt a cautious approach. If you are unsure whether data is safe for input, err on the side of caution and withhold it.

8 of 18

Approved Uses

edit_note

Drafting

Utilize tools to create initial outlines and content structures.

psychology

Brainstorming

Generate creative ideas and explore diverse perspectives quickly.

search

Research

Gather information and summarize complex data points efficiently.

accessibility_new

Accessibility

Improve content reach and inclusivity for all audience members.

Customize this with real examples your staff already do. I have used perplexity to find Edgar award nominated and winning books in one tab, have the public catalog open in another tab and the perplexity computer checks each book in the list on the first tab for inclusion in our catalog in the second tab and created a table that I used to populate an FB post for our system’s page. This is the sort of public use (assisting with research and discovery) that you can maybe pre-approve

Talk to staff about use - I heard at ALA a lot of stories of disconnect between admin and staff at libraries - not always the same sort of disconnect, but one was pro and the other anti and neither one realized the attitude of the other before trying to put together a policy�4. Approved Uses of AI

AI may be used to support (but not replace) staff work, including:

Drafting and editing content (marketing, policies, communications)
Brainstorming ideas and programming concepts
Assisting with research and information discovery
Creating training materials
Enhancing accessibility (e.g., alt-text, summaries)
Supporting patron learning about AI tools

[OPTION: Add local examples specific to your libraries]

9 of 18

Real World Examples

edit_note

“Writing a program description”

description

“Creating a board report draft”

supervised_user_circle

“Helping a patron evaluate AI output”

10 of 18

Prohibited Uses

lock

No confidential data

smart_toy

No automated decisions

visibility_off

No surveillance

11 of 18

Tool Approval

layers

Optional structure

Flexibility in how the tool is organized or integrated into current workflows.

verified_user

Director approval

Formal sign-off required from department leadership before implementation.

assignment_turned_in

Review process

A standardized evaluation path to ensure compliance and technical fitness.

12 of 18

Staff Training

gavel

Ethics

Understanding the moral implications and responsibilities of using AI tools in the workplace.

security

Privacy

Maintaining data confidentiality and protecting patron information in all digital interactions.

psychology

Bias awareness

Recognizing and mitigating algorithmic biases to ensure fair and equitable service delivery.

A policy without training won’t work. Plan how you’ll support staff. what to train staff on? From recent Computers in Libraries: "AI literacy and conceptual understanding require a grounding in basic AI concepts—what AI is, what it isn’t, and how it works at a high level. This includes understanding machine learning (supervised, unsupervised, reinforcement), NLP (how computers interpret and generate human language), neural networks and deep learning, algorithms versus heuristics, and data structures and patterns."

7. Staff Training and Expectations

Staff will receive training on:

Ethical AI use

Privacy and data protection

Identifying bias and misinformation

Effective prompting and evaluation of outputs

Staff are expected to:

Critically evaluate AI outputs for accuracy (“AI hallucinations” are possible)

Use AI appropriately within policy guidelines

Seek guidance when unsure

13 of 18

Patron Education

school

AI Literacy

Empower patrons by teaching foundational skills to navigate and understand AI technologies.

groups

Workshops

Offer structured, interactive sessions focused on practical AI applications and tools.

help_outline

Guidance

Provide one-on-one support and curated resources for ethical and effective AI use.

This is where libraries lead—helping communities understand AI. This is where we should be focusing our attentions right now, too!

8. Patron Use and Education (KEY SECTION)

The library recognizes AI literacy as a core digital literacy skill.

8.1 Patron Access

Patrons may use AI tools provided or permitted by the library

Access will align with existing internet and technology use policies

8.2 Patron Education (EMPHASIS AREA)

The library will actively support patrons in understanding AI, including:

What AI is (and is not)

Strengths and limitations of AI tools

Risks (privacy, misinformation, bias)

How to evaluate AI-generated content

Ethical and responsible use

This may include:

Workshops and classes

One-on-one assistance

Guides and online resources

8.3 Staff Role in Patron AI Use

Staff may:

Demonstrate AI tools

Help patrons craft prompts

Teach evaluation and verification skills

Staff should NOT:

Present AI outputs as authoritative without verification

Input patron personal data into AI systems

14 of 18

Living Document

update

Review regularly

Ensure policy remains relevant and accurate through scheduled, periodic assessments.

edit_note

Adapt as needed

Update content to reflect evolving technology, community needs, and organizational changes.

15 of 18

How to Fill It Out

find_replace

Replace placeholders

Update all bracketed text with your specific organization's details and contact information.

tune

Customize examples

Tailor the policy scenarios and use cases to reflect your local community and library workflows.

visibility

Keep it readable

Ensure formatting remains clean and language is accessible for all staff and board members.

16 of 18

AI in Other Policies

library_books

Collection Development

handshake

Vendor Management

public

Social Media

groups

Patron Behavior

???

17 of 18

Using the Board Summary

send

Share ahead

Provide the summary to board members in advance to allow for thoughtful review.

center_focus_strong

Focus on values

Anchor discussions in core organizational values to ensure alignment and integrity.

school

Highlight education

Use the summary as a tool for ongoing board education and strategic learning.

18 of 18

Closing

task_alt

Focus on Progress

You don’t need perfect. Striving for perfection can stall implementation.

build

Prioritize Utility

You need usable. A functional policy is more valuable than a flawless one.

Robin Hastings

Library Services Consultant, NEKLS

rhastings@nekls.org

https://www.rhastings.net/ai-resources/ - Halfway down the page, under Templates From Robin