Writing an AUP starts with the decisions that make writing the policy possible. 

I’ve been in rooms where a group of leaders is trying to write an AI acceptable use policy, and most of them hit the same wall about 40 minutes in. Someone says we should define what we mean by artificial intelligence first. Then, someone else pulls up an external district’s policy and reads it aloud, quickly shifting the meeting into a group critique of a document no one presently owns or supports. Two hours later, there’s a shared doc with fourteen comments on it, and not one decision has been made.

That is not a competence problem. Those rooms are usually full of intelligent, thoughtful people who genuinely want to get this right. The problem is that a committee was asked to “create a policy” when what they needed to do was make decisions. Those are two entirely different meetings.

I also understand the desire to create a long, all-inclusive policy document. When something feels big, and you need it fast, a lengthy document creates the illusion that everything is covered. If we address every possible situation, we cannot be blamed for missing anything. I’ve felt that pull myself. Years ago, when I was writing a device acceptable use policy, I remember believing that if I just anticipated enough scenarios, I could write something that would never leave decisions up for interpretation. More detail equals better understanding, right? That’s not how any of this works. And it’s certainly not how it works with AI in schools, where the tool your staff is using in November may not even resemble the one you approved in September.

After tons of conversations with K-12 leaders, here are some common challenges with creating your district’s AI acceptable use policy.

The Committee Will Prioritize Wordsmithing Over Decision-Making

Every district writing AI guidelines is facing the same fifteen or twenty questions, and about six of them are genuinely hard. Do we permit anything not specifically prohibited or prohibit anything not specifically approved? Does the policy cover the AI features already sitting inside the tools we approved two years ago? Can a family opt their child out?

Nobody can answer those for you. Not me, not a sample AI policy template from a district three states away, and not an association model policy. They’re value questions disguised as operational decisions. Until your group agrees on the values that will make that determination, there’s nothing to write. Once your group answers them, writing the policy becomes the easy part.

This is why I stopped recommending that policy committees start with a draft. While it can be valuable to review other district AUPs, ultimately, decisions need to be made on how your district’s values will shape that document. Therefore, start with the decisions. Get them on paper and hand them to whoever writes the policy language. That person will thank you, and you’ll save time by avoiding the group wordsmithing session you were bound to have.

Split It Into Two Documents

Your board policy should hold the things that will still be true in three years: the principles, the prohibitions, who is accountable, and who owns the final document. Your administrative regulations should handle everything that changes: the approved AI tool list, the request and vetting process, and grade-band specifics.

If you put your approved tool list in board policy, you’ve just committed to a board vote every time a vendor adds a new feature. I watched a district sit on an approval for six weeks waiting for the next board meeting, while teachers used the tool anyway, or the tool they used for instruction must be shut down until it can be approved. That’s worse than having no policy, because now you’ve taught your staff that the policy doesn’t matter.

Six Sections Is Enough

1. Scope: Defines who the policy applies to and whether it covers AI features baked into tools you’ve already approved.

2. Approved tools: What is on the list, who adds to it, and how the rules adapt by grade level. This is where COPPA compliance and your district’s rules for students under 13 live.

3. The non-negotiables: Student data privacy takes top priority here. Clarify rules on not inputting personally identifiable student information into unapproved tools, using AI-generated likenesses of real students or staff, or relying solely on AI to determine grades, placements, or disciplinary decisions. This section requires your dedicated focus, as it’s where FERPA exposure happens.

4. Disclosure: Students and staff say when they use AI. The teacher sets the permitted level of AI assistance for each assignment within their instruction. This is your academic integrity hook. It should point back to your instructional framework to support the degree to which AI will be utilized.

5. Human accountability: You submit it; you own it, including the parts that are wrong.

6. Ownership and review: A named person and a named date.

That is a policy people will read, which is the entire point. A fourteen-page policy is not necessarily more protective than a two-page policy. It’s less likely to be remembered, harder to follow, and considerably more likely to contradict itself in the middle.

Three Questions That Will Decide If Your Policy Survives

When making the decisions, slow down on the following often overlooked situations that might require a little more intentional thought.

AI detection tools: Decide now whether a detection score is evidence, one signal among several, or nothing at all. AI detection tools have been found to be unreliable, suffering from high false positive rates that aren’t evenly distributed. Multilingual students and students who write in plain, direct sentences get flagged more often. If your policy does not say what weight that number carries, then the weight it carries is whatever the individual teacher decides, and that’s not a system that supports every student.

AI in IEP, 504, and evaluation report drafting: This is where your most sensitive student information has the potential to end up pasted into a chat window. It doesn’t take a dysfunctional teacher to make this mistake. It could be your best special education teacher at nine at night, trying to get through eleven drafts, using the tool that helped her last week. She’s not being careless. She’s under-resourced, and the tool is right there. You need an answer to this question that acknowledges why it happens in the first place.

Family opt-out: If you offer an opt-out, someone must design an alternative assignment, and that someone is a teacher. Setting expectations for teachers, families, and students about what happens if a student is allowed to opt out is a proactive way to ensure that teachers can do their jobs and families’ expectations are reasonable.

Put a Name and Date on It

Most school board policies get reviewed on a cycle measured in years. However, I would argue that AI policy needs to be evaluated twice a year for now, with assigned ownership and the review date printed at the bottom of the document. I know that sounds like a lot. But consider what has changed in the last 18 months and then consider signing your name to a document that claims it doesn’t need another look until 2028.

Where to Start

Below is a free AI policy decision sheet to support this work. It’s four pages with 18 questions, checkboxes, space for notes, and a table showing which of the six policy sections each answer relates to. It’s not a policy template. This resource can guide the decision-making you have before creating a policy, and a district committee can get through it in one sitting.

Your district does not need a longer AI acceptable use policy. It needs a shorter, more practical one that people can actually use.

Download the AI Acceptable Use Policy Template Here!

Need help facilitating this work? Our PD team at Teq can assist with that!