Estimated reading time: 7 minutes
FERPA and AI is one of the first questions a district should ask before adopting any AI tool. FERPA, the Family Educational Rights and Privacy Act, governs how schools handle student education records, and AI tools that process student data fall squarely under it. This guide explains, in plain terms, what FERPA requires of AI in K-12, how the school official exception applies to AI vendors, and how to vet a tool for compliance. It is general information, not legal advice.
Table of contents
Executive Summary
FERPA gives parents rights over their children’s education records and limits how schools disclose personally identifiable information from those records. When an AI tool processes student data, the district remains responsible for FERPA compliance. The most common path for using a vendor is the school official exception, which lets a district share education records with a vendor performing a service the district would otherwise do itself, provided the vendor is under the district’s direct control and uses the data only for that purpose. The practical takeaway: require a FERPA-compliant data agreement before any AI tool touches student data.
Key Takeaways
- FERPA and AI: AI tools processing student records fall under FERPA.
- The district stays responsible for compliance, even when using a vendor.
- The school official exception is the usual basis for vendor access.
- Vendors must be under district control and limit data use to the service.
- Require a data-privacy agreement before any pilot.
What FERPA Requires
Quick answer: FERPA requires schools to protect the privacy of student education records and generally to get consent before disclosing personally identifiable information from them, with specific exceptions. An education record is information directly related to a student and maintained by the school. When AI processes that information, FERPA’s protections travel with the data.
The School Official Exception and AI Vendors
FERPA lets a district disclose education records, without parental consent, to a contractor or vendor that qualifies as a school official with a legitimate educational interest. To rely on this for an AI vendor, the district must show the vendor performs a service the district would otherwise use its own staff for, is under the district’s direct control regarding the data, and uses the records only for the authorized purpose, not for its own ends. A written data agreement is how a district documents these conditions.
What to Require From an AI Vendor
| Requirement | Why it matters under FERPA |
|---|---|
| Written data-privacy agreement | Documents the school official conditions |
| Use limited to the service | FERPA bars repurposing of records |
| No training of public models on student data | Prevents unauthorized use and disclosure |
| District control and deletion rights | Keeps the district in control of records |
FERPA, State Guidance, and District Policy
FERPA is the federal floor, not the whole picture. Many states add their own student-data-privacy laws, and as of 2026 most states publish K-12 AI guidance that reinforces privacy expectations, while a growing number of districts have board-adopted AI policies. Treat FERPA as the baseline and layer your state’s rules and your district’s AI policy on top. See AI Safety in Schools for how this fits the wider safety picture, and COPPA and AI for under-13 students.
Common Mistakes District Leaders Make
- Letting staff use consumer AI with student data and no agreement.
- Assuming a vendor is compliant without a written agreement.
- Ignoring whether data trains public models.
- Overlooking state laws stricter than FERPA.
Recommended Actions
Immediate (this month): Inventory AI tools touching student data and confirm each has a data agreement.
Medium-term (this year): Standardize a FERPA-compliant data-agreement requirement in procurement.
Long-term (strategy): Make privacy review a gate every AI tool passes before adoption.
Questions District Leaders Should Ask
- Is there a written, FERPA-compliant data-privacy agreement?
- Does the vendor use student data only for our service?
- Is student data kept out of public model training?
- Do our state laws add requirements beyond FERPA?
What a FERPA-Compliant AI Data Agreement Should Contain
The written data-privacy agreement is where the school official exception becomes real. A strong agreement for an AI vendor goes beyond a generic privacy policy and spells out specific protections. Look for clauses that confirm the district owns and controls the student data, that the vendor uses it only to provide the agreed service, and that the vendor will not sell it or use it to build or train products for other customers. The agreement should name any subprocessors the vendor relies on and bind them to the same terms, because student data often flows to cloud infrastructure and third-party services behind the scenes.
Two operational clauses matter just as much. First, deletion: the agreement should state that the district can require return or destruction of student data on request and at the end of the contract, which aligns with the district’s obligation to keep records under its control. Second, breach notification: the vendor should commit to notifying the district promptly if data is exposed, so the district can meet its own reporting duties. A short review with district counsel against these points is far cheaper than discovering a gap after a tool is in every classroom.
Where Generative AI Creates New FERPA Questions
Generative AI raises issues that older classroom software did not. The central concern is training: if a tool uses whatever staff or students type into it to improve a public model, then student records could leave the district’s control, which the school official exception does not permit. Ask every vendor, in writing, whether student inputs are used for model training, and require that they are not. Beyond training, be aware that prompts, chat transcripts, and usage logs can themselves become education records when they contain personally identifiable information about a student, so they deserve the same protection as a grade or an assessment score. The safest posture is to give staff approved tools with clear agreements and to instruct them never to enter identifiable student information into consumer AI that has not been vetted.
Frequently Asked Questions
How do FERPA and AI relate?
FERPA protects student education records, and AI tools that process those records fall under it. The district stays responsible for compliance, so any AI vendor handling student data must meet FERPA’s conditions, usually documented in a data-privacy agreement.
Can schools share student data with an AI vendor under FERPA?
Generally yes, through the school official exception, if the vendor performs a service the district would otherwise do, is under the district’s direct control over the data, and uses it only for that purpose. A written agreement documents these conditions.
Is using ChatGPT with student data a FERPA risk?
It can be, if staff enter personally identifiable student records into a consumer tool without a data agreement and controls. Districts should provide approved tools with proper agreements and set clear staff guidance.
Is this legal advice?
No. This is general information to help district leaders ask the right questions. Consult your district counsel and your state’s student-privacy office for decisions about specific tools.
What clauses should a FERPA data agreement include?
Confirm district ownership and control of the data, use limited to the agreed service, no sale or use for training other products, named subprocessors bound to the same terms, deletion or return rights on request, and prompt breach notification. These clauses document the school official exception conditions.
Can an AI vendor train its models on our student data?
Not without breaking the school official exception. If a vendor uses student inputs to train a public model, records leave the district’s control. Require in writing that student data is never used for model training and is used only to deliver your service.
Are AI chat transcripts education records?
They can be. When prompts, transcripts, or usage logs contain personally identifiable information about a student, they meet the definition of an education record and deserve the same FERPA protection as any grade or assessment result.
Related Resources
- AI in Education: Complete Guide
- AI Safety in Schools
- COPPA and AI
- Best AI Tools for Schools
- AI for English Learners
Conclusion
FERPA and AI come together the moment an AI tool touches student records: the district stays responsible, and the school official exception is the usual basis for vendor access, backed by a written data agreement. Layer your state’s privacy laws and district AI policy on top of that federal floor. The single most important habit is simple: no AI tool touches student data without a FERPA-compliant agreement in place.
Sources: U.S. Department of Education, Student Privacy Policy Office; FERPA regulations, 34 CFR Part 99.
This article is general information, not legal advice. Consult your district counsel for guidance on specific tools.
When the Agreement Is the Easy Part
A district can require a data-privacy agreement for every AI vendor and still find, months later, that staff have been pasting student names into a consumer chatbot that never signed anything. The paperwork was in place; the practice quietly went around it.
The difficulty is structural. Consumer AI was not designed for the school official exception, so a district spends its energy bolting FERPA conditions onto products that assume an anonymous adult user and often reserve the right to train on whatever is typed in.
Districts on firmer ground provide approved tools that meet FERPA by design and make the safe option the easy one. Telo AI is one example built for K-12, with a data-privacy agreement, district control over records, and student data kept out of public model training.
See how a K-12-native tool is built around FERPA requirements: https://mytelo.ai/how-telo-works/
A signed agreement is necessary, but it is not the finish line. The districts that stay compliant are the ones that shrink the risk their agreements have to cover, by giving staff tools that were safe to hand student data in the first place.