I use AI in my writing process. I can ask it to find something I wrote months ago, connect ideas across my work, or help me locate an argument buried somewhere in my folders. For someone who has spent years writing, that’s useful. My filing system has never won any awards.
During my conversation with Steve Smith, founder of the Student Data Privacy Consortium, and student privacy expert Amelia Vance, Amelia made a distinction that deserves more attention.
AI works for her partly because she already knows the subject. She can recognize an answer that looks convincing but is wrong.
That matters when we move from an adult finding a tool useful to a district deciding whether students should use it. We’ve changed the user, the context, and potentially the consequences. Our review needs to account for that.
The familiar vendor has a new feature
Steve brought this down to a problem technology directors are facing right now: AI is becoming part of applications districts already use.
You may know the vendor. You may have a data privacy agreement. Your teachers may have used the product for years.
Then an AI feature appears.
Steve’s point was that districts still need to ask how that feature operates. What information does it receive? Does another company process it? How is that information used and protected?
The agreement remains important. So does understanding whether it addresses what the product now does.
That means approval needs a way to stay current. A new feature or use may warrant another conversation with the vendor—ideally with someone who can explain the data practices, rather than send you a brochure containing the word “responsible” seventeen times.
Privacy is only part of the review
Steve distinguished among privacy, security, and safety. Those questions overlap, but they aren’t interchangeable.
Understanding where information goes doesn’t fully answer how an AI system might interact with a student. A tool can raise concerns about the advice it gives, the relationships it encourages, or what happens during an extended conversation.
Amelia pushed us to take those possibilities seriously, including when students begin with an ordinary task such as homework.
For a technology director, this complicates the familiar approval process. A tool’s purpose on paper may not describe every interaction it enables. Reviewing a teacher’s administrative use and reviewing direct student use can require different questions.
We need to be specific about who will use the tool, for what purpose, and with what supervision. Otherwise, “approved” ends up carrying more reassurance than the review can support.
Enthusiasm needs an explanation
I remain excited about what AI could make possible in education. During the conversation, I described how it helps me work with my own writing. Amelia also described uses she finds valuable.
Her challenge was about what districts can explain and demonstrate today.
A parent asking why their child is using a product deserves more than a prediction about the future of personalized learning. They deserve to understand the activity, its value, and the precautions around it.
That is a fair expectation. It should also influence what we ask vendors to demonstrate.
If we can’t explain why a tool belongs in a particular learning experience, adding a privacy review doesn’t supply the missing educational purpose.
Your approval process needs people who understand it
When we discussed who should own these decisions, Steve returned to district culture.
You can establish review procedures and technical controls. But staff still need to understand why those procedures exist. Otherwise, every new application becomes another contest between someone trying to get work done and someone trying to prevent a problem.
Technology leaders cannot personally supervise every decision. They need instructional colleagues who can assess educational value, teachers who understand when to request a review, and leadership that supports the process.
My takeaway is that an approval should describe a decision clearly enough to revisit it: the intended use, the information involved, the conditions, and who is responsible when something changes.
AI makes that ongoing work more urgent. The vendor’s name may stay the same. The district’s understanding of the product has to keep up.
