Can teachers use ChatGPT with student data? A privacy threat model for educators

The Morse teamPublished November 9, 202612 min read

Teachers can use ChatGPT with student data only in a version their school or district has authorized to handle it, such as a ChatGPT for Teachers or ChatGPT Edu workspace covered by a student data privacy agreement, and never in a personal ChatGPT account. As of November 2026, FERPA lets a school share students' education records with an AI company without a parent's consent only when that company works "under the direct control" of the school, and a personal account carries no such agreement. Your own data is a separate risk: a grievance, a job search, or a health question typed into a school account or on a district laptop can be read by your district, and at a public school, work chats may be public records.

On November 9, 2026, we read the FERPA and IDEA regulations, the Department of Education's guidance on online services and its 2026 AI grant priority, the FTC's amended COPPA rule and its school FAQ, California's student privacy law, Washington State Archives' AI guidance, and the teacher and education terms of OpenAI, Anthropic, Google, and Microsoft. OpenAI and Anthropic deserve credit here: both publish their teacher terms and student data agreements in full, so a teacher can read exactly what the "Agree" button commits them to. This isn't legal advice. Your district's AI policy and its privacy office come first.

Is ChatGPT FERPA compliant?

The personal version isn't, and the education versions can be, depending on who signs. FERPA's regulations let a school treat an outside company as a "school official" only if it does work the school would otherwise use employees for, "is under the direct control of the agency or institution with respect to the use and maintenance of education records," and follows FERPA's limits on passing records on (eCFR, 34 CFR 99.31). A Free, Plus, or Pro account has none of that. On those plans, OpenAI says, "we may use your content to train our models" unless you turn off Improve the model for everyone under Settings > Data controls (OpenAI).

The education plans come with an agreement. OpenAI's Student Data Privacy Agreement, which covers ChatGPT Edu and ChatGPT for Teachers, says OpenAI "shall be considered a School Official with a legitimate educational interest" and "shall be under the direct control and supervision of Customer" (OpenAI Student DPA). OpenAI's help page describes ChatGPT for Teachers as having "FERPA-aligned protections," and says what you share there "is not used to train our models by default" (OpenAI).

In a ChatGPT for Teachers workspace you set up yourself, the "Customer" in that agreement is you. OpenAI's Teacher Access Terms have you "represent and warrant" that "you have received all authorizations necessary from the school, school district, or other similar educational institution" to "accept these Terms on behalf of your School" (OpenAI). Anthropic's K-12 terms for Claude for Teachers say you "may not enter into these Terms on behalf of a School unless you have the legal authority to bind the School," and that the customer will defend Anthropic against third-party claims related to its "Inputs" (Anthropic). A public sector addendum drops that duty where state law bars it, but only when the customer is a public school or district itself, not a teacher who signed up alone. The school official arrangement holds only if your school actually authorized it. A teacher who clicks through without that authorization has vouched for something that isn't true.

The Department of Education warned about this in 2014. With click-wrap terms, it said, "the act of clicking a button to accept the TOS serves to enter the provider and the end-user (in this case, the school or district) into a contractual relationship," and "individual teachers may not understand the specifics of how the provider will use and secure student data" (US Department of Education). Anthropic edited its own launch post on August 28, 2026. The post had shown teachers analyzing student data; the update says "using identifiable student information requires authorization from the school or district" and swaps in de-identified examples (Anthropic).

What counts as student data in a teacher's prompt?

More than names. FERPA protects personally identifiable information from education records, including details that would let "a reasonable person in the school community, who does not have personal knowledge of the relevant circumstances," identify the student "with reasonable certainty" (eCFR, 34 CFR 99.3). "The only girl on the robotics team, who was suspended last week" can pass that test with no name in it. OpenAI's agreement lists what education records can hold: "disciplinary status, test protocols and individualized education programs," among others.

Three kinds of teacher material need the most care:

• IEPs and 504 plans. A 504 plan your school keeps fits FERPA's definition of an education record, and IEPs carry a second layer of rules under IDEA. IDEA asks each participating agency to train everyone who uses the information and to keep "for public inspection, a current listing of the names and positions of those employees within the agency who may have access" to it (eCFR, 34 CFR 300.623). A chatbot account you opened yourself sits outside all of that.
• Parent emails. A parent's message about their child, kept in your school mailbox, is directly related to a student and kept by the school. Pasting it into a chatbot to draft your reply discloses it.
• Your own notes. FERPA leaves out records "kept in the sole possession of the maker," used "only as a personal memory aid," and "not accessible or revealed to any other person except a temporary substitute." Notes typed into an AI company's servers have been revealed to that company. We found no Department of Education guidance on whether they keep the exception after that, so assume they don't.

What do the teacher AI plans promise?

All four big education offerings say they don't train on your chats. They differ on who signs and how much your school can see. Read on November 9, 2026:

Google's admin help adds that a school chooses to keep Gemini chats for 3, 18, or 36 months, and even with history off, Google "may still store these conversations for up to 72 hours" (Google). Microsoft says Copilot's web searches go to Bing, where Microsoft "acts as an independent data controller" under its consumer agreement, though the queries are sent "with user and tenant identifiers removed."

Can your school see your ChatGPT for Teachers chats?

Yes, or it can later. OpenAI's Teacher Access Terms say your account "may later be moved into or associated with your School's admin-managed workspace," after which your administrator can access your content and "restrict or remove your access to the account" (OpenAI). A district's domain claim "can direct people with matching email addresses to the claimed workspace," OpenAI's help page adds. Claude for Teachers works the same way: "Through domain capture, teachers who already verified on the school or district's domain move into the centrally managed account" (Claude).

OpenAI's terms tell teachers, "Because your School can take control of, and become an Admin to your account, you should only use the Services and your Teacher Account for educational purposes." Your teacher account "is separate from your personal OpenAI account." That advice fits every school account. A draft grievance, a resignation letter, or a question about your own health doesn't belong in one. If you're organizing with colleagues, AI privacy for union organizers covers what an employer can reach.

Can your district see a personal ChatGPT account on a school laptop?

It can be set up to. Microsoft's Purview tools can capture prompts and responses sent to ChatGPT, Google Gemini, Microsoft's consumer chat, and DeepSeek in the Edge browser on devices an organization has enrolled. The prompts are "stored in a user's mailbox" and can be searched in an eDiscovery case (Microsoft Learn). Whether your district turns this on is up to your district, and its acceptable use policy is where it would say so. We cover the device layer, app by app, in can my employer see my ChatGPT.

Are a teacher's AI chats public records?

At a public school or university, the work ones can be. Washington State Archives, in guidance updated November 2026, says a generative AI interaction, "input and output," that relates to public business is a public record, and "Agency work-related generative AI interactions created or received using a personal device/account are still public records" (Washington State Archives). Other states haven't all said so, and can government employees use ChatGPT compares the rules we found.

People already ask for these chats. On November 3, 2025, a writer at Defector asked the University of Oregon for "Chat histories of all ChatGPT or Microsoft CoPilot sessions conducted by head football coach on school-owned devices." The university closed the request on November 19, 2025 with "No Responsive Records" (University of Oregon).

Can teachers use AI tools with students under 13?

Only tools your school has vetted, and none of the teacher plans is for students. ChatGPT for Teachers "is not for students" (OpenAI), and Claude for Teachers is "for educators only, consistent with Claude's 18-and-over policy" (Anthropic).

For apps that collect information from children under 13, COPPA lets a school consent for parents only "in the educational context," where the company collects it "for the use and benefit of the school, and for no other commercial purpose" (FTC). The FTC's amended COPPA rule took effect June 23, 2025, with most of it binding from April 22, 2026. The FTC left out its proposed school rules "to avoid making amendments to the COPPA Rule that may conflict with potential amendments to DOE's FERPA regulations," and says it will keep enforcing in schools "consistent with its existing guidance" (Federal Register).

State laws add their own limits. California's SOPIPA covers information "created or provided by an employee or agent of the school" to a service built for K-12 school purposes, but it "does not apply to general audience internet websites" (Cal. Bus. & Prof. Code 22584). Student details a teacher types into a general chatbot fall outside it.

Does the Department of Education have AI rules for teachers?

No new ones on privacy. Its final priority on AI in education, published April 13, 2026 and effective May 13, steers discretionary grants toward AI projects. Commenters asked for privacy safeguards and parental notice, and the Department answered that "how best to ensure safety and communicate about technology use is optimally decided at the state and local level" and "declines to enact requirements at the federal level" (Federal Register).

That leaves FERPA, COPPA, your state's law, and your district's contracts. OpenAI said on August 26, 2026 that it has signed statewide offers through the Student Data Privacy Consortium in Illinois, Iowa, Maine, Massachusetts, Missouri, Nebraska, New Hampshire, New Jersey, Ohio, Rhode Island, Tennessee, Texas, Vermont, Virginia, and Washington, plus a separate agreement covering California (OpenAI). In those states, your district has a ready-made path to approve ChatGPT for Teachers. Ask whether it has taken it.

Who could see a teacher's AI chats: the threat-to-tool table

Built from the sources above, read November 9, 2026.

How Morse keeps a teacher's own questions off school systems

Every row in that table runs through a copy of your chat that someone else keeps: an AI company's servers, your district's Vault or mailbox, or an admin console that can absorb your teacher account. Student records belong in the tools your district has approved, under its agreement. Your own questions need a place outside those systems. Morse stores your chat history only in your browser, encrypted, and our servers don't keep your messages or log what you write.

Use it on your own phone or computer, signed in with a personal email, since sign-in is a code sent to that inbox. Then there's no district account for an admin to claim, and no school mailbox for a Purview search. Pick a mode per chat:

• E2EE, for what you'd never type into a school account. A grievance, a complaint about an administrator, a lawyer's letter, or your own health. Your browser checks the secure enclave's proof that it's genuine before anything leaves your device, then encrypts the message so only that enclave can read it, and Morse's servers relay text they can't read. That rests on the enclave hardware and attestation working as designed. E2EE runs open-weight models, without web search or image generation.
• Private, for lesson work with no student in it. A unit plan, a rubric, a reading list, or a parent newsletter with placeholders for student names that you fill in later in the district's system. Private is the default for a new chat, and Morse has zero data retention agreements with its inference, search, and other providers, so they keep nothing.
• Anonymous, to ask GPT, Claude, or Gemini about your contract, your license, or a new job without your name on it. Our proxy removes your account, email, and IP address and sends the message under Morse's own account, so the provider sees the question without knowing it's you. It may keep the message for a limited time to check for abuse, under its own policy, so leave out your name, your school, and your district.

In every mode, Morse doesn't train AI on your content, sell your data, or show ads (privacy policy). History locks when you sign out, and unlocking it takes two keys, one of which never leaves your device. Because we don't store chats, we usually have very little to give authorities. Clearing your browser deletes your history for good, and Morse can't recover it, so keep a copy of anything you'll need. Morse is for adults: you confirm you're 18 or older. For a job search, private AI for job seekers covers what hiring tools keep.

Where each kind of work goes:

• Anything with a student in it: the AI plan your district authorized, or no AI at all.
• Lessons, rubrics, and templates with no student in them: any tool you like. On Morse, Private.
• Your own life, on your own device: on Morse, E2EE for the sensitive questions, Anonymous to ask GPT, Claude, or Gemini.

Try a chat from the Morse home page without an account, or start free with no card needed. End-to-end encrypted chats come with Pro, at $19 a month.