FERPA Questions for Reviewers of AI Tools That Handle Student Data
AI tools reach student data in more ways than a chatbot. Students may type questions into a course assistant, faculty may run submissions through a grading or feedback tool, and analytics tools may pull records straight from the LMS. For the people who approve software at a university, the first question is usually what happens to that information. In the U.S., that question runs through the Family Educational Rights and Privacy Act, known as FERPA.
As someone who builds AI tools for higher education, I would expect a review to follow student data through the whole tool. The model provider's contract is one part of that review, and the integration, storage, and services around the model matter just as much. (I wrote more about that larger system in Beyond the Model: AI Privacy, Safety, and Security for Students.)
This post covers how FERPA treats outside vendors and which questions are worth asking. I am not a lawyer and this is not legal advice. Your privacy office and counsel make the call for your institution.
FERPA Basics for AI Tool Reviewers
FERPA generally requires consent before a school discloses personally identifiable information from a student's education records. There are exceptions. One of them is often called the school official exception. The U.S. Department of Education's student privacy site explains that a school may outsource institutional services or functions that involve disclosing education records to a contractor or other third party if that party meets four conditions. It performs a service the institution would otherwise use its own employees for. It is under the direct control of the institution when it comes to the use and maintenance of the records. It uses the information only for the purposes for which it was disclosed and follows the limits on passing it on. And it meets the criteria the institution has set out in its annual notice of FERPA rights for school officials with a legitimate educational interest.
That is a lot of text, and the useful reading of it is simple: the school stays in charge, and the vendor can only use student information for the job it was given.
FERPA Questions to Ask an AI Vendor
One question comes before any of the FERPA conditions: what does the tool collect? Ask which LMS or roster fields it pulls and what each one is used for. Many tools can do their job with far less than they request. A course assistant, for example, may only need the student's question and the relevant course material. Data the vendor never receives is data you never have to protect, audit, or delete from their system.
From there, take each of the four conditions and turn it into a question. The vendor does not need to use the word FERPA in their answer, but their answer should make it clear that they can meet the condition.
Start by asking who owns the data, and whether the institution can have it deleted or returned when the contract ends. Retention deserves its own question, because some records must be kept for years in case a complaint, investigation, or legal matter later requires access to them, so find out how long the vendor keeps student records and audit logs and whether that period can be matched to your institution’s records retention schedule. When it comes to purpose, ask whether student interactions are used to train or improve models, and if so, whether you can opt out in the agreement. Student data may also pass to other companies, such as model providers or cloud hosts, so ask which ones receive it and what terms bind them. Finally, check with your registrar or privacy office about whether this type of vendor fits the school’s current definition of a school official.
One thing to note: using a student's question or draft as context for a response is different from using it to train a model. Ask the vendor to explain which use applies. If the agreement promises no training on student data, does that commitment cover the vendor and every model provider it uses?
One state has started to settle the training question. California's AB 1159 was signed on September 10, 2026. Its higher-education protections begin July 1, 2027 and prohibit covered education-technology providers from using identifiable student data to train generative AI or develop other AI systems. Even outside California, it's reasonable to ask vendors for the same commitment in writing.
Questions to Ask a Vendor About Data Storage and Retention
Beyond FERPA, reviewers usually want to know where data is stored, how long it is kept, how access is limited, what happens after a security incident, and how accessibility is documented. EDUCAUSE has written about how institutions are approaching AI procurement, including concern about faculty and staff entering student or institutional data into tools that have no agreement in place. That concern is a good reminder that an approved tool can be a better choice than a free one picked up on the side, because there is a contract behind it.
The main database is rarely the only copy. Search features often store student writing in a converted form called embeddings, error logs can record full conversations, and outside monitoring services may receive copies as well. Ask the vendor to list every location where student data is kept, the retention period for each, and how deletion works across all of them.
Whatever you ask, get the answers in a document you can file. A vendor who tells you something on a call has told you very little. A vendor who puts it in a data processing agreement, a privacy policy, or a security document has told you something you can check later. If a document does not exist yet, that is worth knowing, and a good vendor will say so.
Notifying Students About AI Tool Data Use
Whatever the legal answer, students should hear a plain explanation before they use the tool: what is collected, who at the university can see it, whether it is shared with any other company, and who to ask if they have a concern. A short paragraph in the syllabus or at the first login is usually enough. Students are more comfortable with a tool when they can see that the school thought about this before they did.
It also helps to keep a simple rule for your own team: do not paste real student records into any AI tool that has not been through review. That rule is easy to remember and easy to explain.
Nuubi and Student Data Questions
I build Nuubi, an AI-integrated tool for helping students practice using AI in their coursework, inside learning management systems such as Canvas, Moodle, Blackboard, D2L Brightspace, or on its own. Its AI assistant, Cuubi, answers student questions using the course's own content as context (never training). We expect reviewers to ask every question in this post, and we would rather answer them early and plainly than have a team find a gap late. If your institution is reviewing Nuubi or building its own vendor checklist, contact us and we'll walk through our answers.
Related reading: Beyond the Model: AI Privacy, Safety, and Security for Students · Questions to Ask Before Choosing an AI Tool for a University Course
Frequently Asked Questions About FERPA and AI Vendors
Is an AI vendor a school official under FERPA? It can be, if it meets the four conditions of the school official exception, including staying under the institution's direct control and using student data only for the service it was hired to provide. Whether a specific vendor qualifies also depends on how your institution defines school officials in its annual FERPA notice.
Can AI vendors train models on student data under FERPA? FERPA limits vendors to using education records for the purpose they were disclosed for, and training a model is generally a different purpose from providing the service. Reviewers should get the vendor's position on training in writing. California's higher-education protections under AB 1159 prohibit using identifiable student data for AI training or development starting July 1, 2027.
Do students need to be told when an AI tool uses their data? FERPA doesn't always require separate consent when a vendor qualifies as a school official, but students should still get a plain explanation of what is collected and who can see it. A short note in the syllabus or at first login is usually enough. A vendor’s privacy policy is usually publicly available on its website.
Sources
U.S. Department of Education, Protecting Student Privacy: Who is a school official under FERPA?
U.S. Department of Education, Responsibilities of Third-Party Service Providers under FERPA
EDUCAUSE Review, AI Procurement in Higher Education: Benefits and Risks of Emerging Tools (March 2025)
Echelman, A. (2026, September 11). As tech firms collect and sell student data, new state law aims to protect their privacy. CalMatters.
Privacy Rights Clearinghouse. (2026, September 10). Governor signs PRC-Sponsored AB 1159 into law, strengthening privacy protections for students across California.