David Lee King

*WordPress let me insert an AI image! I used the prompt “a person writing a policy about AI tools”*

I’ve been doing some research on AI policies for libraries (well, public libraries, anyway!), because at some point, my library will probably discuss whether or not we should have an AI policy. And we’ve been experimenting with using AI tools for a variety of things.

So – time to start reading some AI policies! I have found a handful of public library AI Policies (linked at the end of this blog post), and I’m finding some interesting things to think about. Here are some observations on what I’m seeing.

AI Policies are all over the place
This is probably similar to other library policies. Some policies tend to cover specifics – what that specific library needs, or is worried about, at the time of creating the policy.

The AI policies I found cover a lot of ground – from approved/not-approved use (including up to potential termination!), to setting up a small AI Approved Tools committee, to the fact that the library needs to train staff in use of new AI tools. Some policies are more positive, while some seem to have a more negative tone in the policy.

Approving specific AI tools does not sound like an easy task to me! Some libraries have designated a staff member or committee that approves which AI tools staff can and cannot use. On the one hand, that’s good, general practice for software-based tools.

On the other hand … that sounds a bit like playing Whack-A-Mole, since so many software and online tools are starting to incorporate AI in their functionality. Do these committees need to approve the use of clicking the better word or phrase in my email when that appears (that’s AI in action)? How about that Facebook/YouTube feed (again, AI in action)? Or all the AI tools that Canva recently released into their system? Depending on what’s really meant… reviewing and approving AI tools could be someone’s full-time job!

Security Concerns Abound
Most of the policies I saw mentioned security in some form. That’s valid – AI tools are sort of the wild west at the moment! I’ll also guess there’s probably a good balance between being heavy-handed vs letting staff members do anything they want.

Ethics/Bias/Fairness/Privacy Concerns
Many of the policies I looked at also mentioned ethical use/bias/fairness/privacy. The ethics sections usually mention bias and fairness issues that can appear in generative AI tools. So a policy might mention not using AI tools to do things like narrow down potential hiring candidates, or the policy might state that all patrons have equal access to services and resources (including AI tools).

Well yes, I agree with that stuff! But my guess is that most of these libraries already address those types of concerns in other library policies. So I’m not sure it needs to be reframed in a separate “oh yeah, here too!” policy. There’s probably a better way to reference those very valid concerns in a way that doesn’t cause the board to have to update multiple policies over one change some time down the road.

Transparency Concerns
Can a library use AI tools to create content for the library? And if so, do you need to somehow announce that you did this? That’s the main concern here. Library AI policies touch upon this idea in a variety of ways. One library policy said no to most AI image-creation tools because of potential intellectual theft when using these tools. They only allow using AI image manipulation tools for what I’ll call “minor edits” – things like adjusting color or saturation of an image. Other libraries specify what types of imagery can be created using these tools (i.e., generic imagery, for use on signs, posters, and marketing material.).

Some library AI policies state that attribution is required when an AI tool is used to create content (image or text). To me, that seems like a strange requirement, if it’s an across-the-board policy. For example, with text-based AI content – if I ask ChatGPT for a better way to word something, I’d need to somehow provide attribution for that (I guess in the form of a citation? Not sure).

And that’s sorta weird. I’ve written a lot of articles and a couple of books. Once I send my manuscript off to the publisher, editors and proofreaders change my words! They keep my meaning, but might provide a better, more succinct sentence. In this instance, the AI tool is doing the same thing. No author attributes that type of editing… (except in a “thank you” section of the book). Same thing here – I don’t think it’s needed.

Same way with images. Sure – if you’re using an AI-created image, you might need to mention that (like I did with this blog post). But if you use an AI image tool to add more background to an image, or to remove things in the background of the image … or to help touch-up an image? That’s no different than what graphic designers have done for years, with or without AI tools. Again, in my opinion – that does not need to be mentioned. That’s just business as usual, just with a newer tool.

Training for Staff and for the Community
This last one is sorta cool. Some libraries are focusing on their commitment to train staff and their community in the use of AI tools in their AI policies. I think that’s great! They are formally stating that they will provide guidance and resources to support their community in figuring out AI and that staff need to be trained, and even need to evolve to use these newer AI-based tools. That’s commitment!

So … do I think libraries need to add an AI policy? Maybe yes, maybe no. Every community is different, so your mileage may vary. It seems like at least in some instances, your other policies probably cover some of what you might want to state in an AI policy. In that case, perhaps a set of guidelines for staff to follow might work better, at least for starters.

If you do decide on a policy – hope this article helps! Also, here’s an article by Nick Tanzi on library I policies and what to think about.

Public Library AI policies: