Education & Student Safety AI
GPTZero
High riskStandalone AI-writing detector with a free web checker and paid tiers, widely used by individual teachers and professors to screen student work for AI generation.
Is GPTZero safe for confidential data?
GPTZero carries two distinct risks and most schools have only thought about one. The visible risk is accuracy: independent testing has reported false-positive rates on student writing far above the vendor's claims (studies and press reports commonly cite figures around 18-20%, versus roughly 1-2% claimed), the 2023 Stanford study found AI detectors overall misclassified 61.3% of non-native English speakers' TOEFL essays, and press reports describe a Yale School of Management student who sued in 2025 after a year-long suspension that relied in part on GPTZero results. The quieter risk is procedural: GPTZero is overwhelmingly used ad hoc — a teacher pasting or uploading a student's essay into a third-party web service with no DPA, no district agreement, and no notice to anyone. That upload is itself a disclosure of a student's education record, so the FERPA question attaches to the teacher's use of the detector, not just to the accusation that follows.
Risk by plan
The same product often carries very different terms depending on the tier — consumer plans are where the exposure concentrates.
The tier where nearly all classroom use happens — no agreement, no admin oversight, and no record of what student work was uploaded.
Higher scan limits and reporting for individual subscribers; still an individual account, not an institutional agreement.
Org plans with LMS integrations exist and are the only route to contractual terms a district can actually review.
Data handling
Training on inputs
GPTZero's public materials state uploaded documents are not used to train its detection models; verify the current terms for the tier in use, since ad-hoc users rarely review them. Anything pasted in is still transmitted to and processed by a third party.
Retention
Scan history is retained for account holders; the vendor does not publish detailed retention schedules for free-tier scans. Assume uploaded student work persists unless deleted.
Residency
U.S.-hosted; no education-specific residency options published. For Canadian teachers, pasting identifiable student work into it is a cross-border disclosure that would normally require a PIA in BC public schools and sits squarely in the conduct the October 2025 FPT commissioners' edtech resolution addresses.
Compliance
- SOC 2Not verified
- GDPR / DPANot verified
- HIPAA BAANo
Certifications typically apply to specific tiers and contracts — confirm scope in writing before relying on them.
New to these frameworks? See our plain-language guides to SOC 2 and the other AI compliance standards.
Enterprise controls
- Organization accounts with shared dashboards
- LMS integrations on education plans
- Writing-report exports for review workflows
Frequently asked questions
How accurate is GPTZero — does it give false positives?
Yes, at rates that matter. Independent evaluations have reported false-positive rates on human-written student work in the high teens to around 20% depending on methodology, and the Stanford TOEFL study found detectors as a class flagged 61.3% of non-native English speakers' essays as AI-written (GPTZero later published de-biasing work against that dataset). No systematic review to date has found any detector reliable enough to serve as sole evidence of misconduct.
Can a student be disciplined based on a GPTZero score?
Institutions have tried, and it is generating litigation: press reports describe a Yale School of Management student suing in 2025 over a suspension based partly on GPTZero results, arguing among other things that detectors discriminate against non-native English writers. Any discipline process should require corroborating evidence — drafts, version history, a conversation with the student — because a probability score from a third-party web tool does not survive scrutiny on its own.
Is it a FERPA problem for a teacher to paste student essays into GPTZero?
It is at minimum a FERPA question the district should be able to answer. A student essay tied to an identifiable student is an education record, and uploading it to a vendor with no DPA or School Official designation is a disclosure the district has not authorized — exactly the shadow-AI pattern of individual adoption outside any agreement. Districts that want detection should procure it institutionally; teachers acting alone should at least strip identifying information, though that does not fully cure the problem.
Policy changelog
- Initial entry published from GPTZero's published materials and cited independent research and coverage.
Sources
- GPTZero
- University of San Diego LRC guide on AI-detector false positives
- Vanderbilt guidance on AI detection reliability
This profile summarizes the vendor's published policies as of the verification date. It is not legal advice.
GPTZero is probably already in your organization.
Sanitized AI shows you who is using it and redacts sensitive data from prompts before it leaves your control.
More AI tool profiles
Chrome extension for teachers that layers AI onto Google Workspace and the sites they already use — generating feedback on student documents, creating materials, and replaying how a student's draft was written.
Consumer companion-chatbot platform where users chat with user-created AI personas; it has no education product, but is one of the most heavily used AI apps among teenagers.
District safety-monitoring service that uses AI plus human reviewers to scan student email, documents, and files in Google Workspace and Microsoft 365 for signs of self-harm, violence, and abuse.
Khan Academy's AI tutor and teaching assistant: a student-facing guide that coaches rather than answers, plus teacher tools, offered to families and through district partnerships.