Generative AI and Museum Publishing: Crafting Policy

April 21, 2026

Museum Publishing Digital Interest Group

Agenda

1. Welcome
2. Survey results
3. Key topics for discussion
4. Introductions
5. Questions from the audience
6. Further reading/Resource library

2

Survey Results

Survey Results: Policies and Contracts

Survey Results:

Usage and Familiarity

Plus: 11 responses with variants of “I haven’t / won’t use it!”

Survey Results:

Fears and Opportunities

Questions museum publishers are facing

Who should participate in crafting policy and getting buy-in?

Should we resist the use of genAI or try to manage it responsibly?

What are our obligations as stewards of creativity, knowledge, and community?

How can we develop AI literacy in our field and institutions?

What are legal concerns or risks related to genAI?

Today’s panelists

Rob Stein

Chief Information Officer

National Gallery of Art

​

Katherine Boller

Editorial Director of Art and Architecture

Yale University Press

​

Allee Austin

Associate Archivist

Amon Carter Museum

​

Katie Reilly

Associate VP of Publishing

Art Institute of Chicago

Supporting AI Literacy & Policy at the National Gallery of Art

Robert Stein, Chief Information Officer

April 21, 2026

AI Policy

• The National Gallery’s first Generative AI Acceptable Use Policy Circular (65-5) was released on June 13, 2024.
• Establishes CIO Authority in AI and integrates with IT policy.
• Based on OMB M-24-10*.
• Defines the difference between Enterprise AI and Commercial AI.
• Establishes an application process for Enterprise AI and a Use Case Registry.
• Currently states that no unsupervised use is allowed
• Currently states that no AI generated images that purport to be works of art are allowed.

Initial Art & AI Workshop hosted by the National Gallery in November 2023.

https://www.whitehouse.gov/wp-content/uploads/2024/03/M-24-10-Advancing-Governance-Innovation-and-Risk-Management-for-Agency-Use-of-Artificial-Intelligence.pdf

Enabling AI Literacy at the National Gallery

• We’ve distilled the 10-page Policy Circular down to a few key communication strategies for our staff.
• Quick Reference Guide on GenAI
• Artificial Intelligence Resource Pages on the Intranet
• In-person trainings offered by the CIO
• Specific training for IT Helpdesk Analysts
• GenAI is a special topic for IT Security training this year
• Community of Practice for Enterprise GenAI
• Launch of NEXTNOW – multi-part AI training and literacy campaign

Rob Stein offering a training on GenAI to Visitor Experience Staff

Commercial AI Tools vs�Enterprise AI Tools…

1. Public Information is information that any member of the public could find easily on the Internet.
2. Public Information may be used with any Commercial AI tool
3. All other data is non-public.
4. Non-public data must be used inside the National Gallery’s approved Enterprise AI tools (M365 Copilot & OpenAI ChatGPT)
5. Custom AI tools may only be built under the supervision of the Digital Solutions Division
6. Third-party AI tools may be approved through procurement by providing requested information to Digital Solutions for review and approval.

Generative AI Acceptable Use Policy

13

DO:

• Review all output from GenAI tools prior to using it.
• Disclose when a GenAI tool was used to create text or content for the public.
• Follow all existing procedures for review and approval of your work whether or not you used GenAI to assist.
• Apply for access to Enterprise GenAI tools when you have a use case you feel is warranted and approved by your Division Head.
• Get permission from your Division Head and the CIO prior to sending National Gallery content to a third-party who will use a GenAI tool to process that content.
• Request that vendors disclose when GenAI tools are used to perform their work and receive approval from your Division Head and the CIO prior to using GenAI generated content.

Generative AI Acceptable Use Policy

14

DO NOT:

• Use any Commercial GenAI tools with National Gallery data not available to the general public.
• Enter any personal information or non-public data into any Commercial GenAI tools. 
• Use your National Gallery email address as the login to any Commercial GenAI tools.
• Use any content from a GenAI tool you suspect may include copyrighted materials.
• Do not use GenAI if you think it might impact the safety of people or property. If you’re unsure, please ask the CIO.
• Do not use GenAI if you think it might impact people’s rights because of implicit bias. If you’re unsure, please ask the CIO.
• Attempt to install or configure any GenAI software or source code onto National Gallery systems without approval and assistance from the Office of the CIO.

1. Limit the types of questions (input) that users can ask the AI.
2. Add caveats and context that content created by AI is potentially biased and unreliable.
3. Use AI-generated content behind-the-scenes but don’t display it to the user.

Don’t use unsupervised AI at all.

Only use unsupervised AI for internal purposes.

Stop worrying so much and pursue scale quickly.

Use unsupervised AI but only behind-the-scenes

Use unsupervised AI but add caveats and context

1

2

3

4

5

Our AI Policy Today

Rob’s gut-feeling on where to land…

Where do you think we should land?

When should we consider using unsupervised AI?

What could we do to mitigate the risks?

Any questions?

Ask away!

16

Resources and readings

See the Generative AI in Museum Publishing Resource Library

Thank you for attending!