Voice-AI in school only works if families and schools trust it. We earn that trust through clear commitments: what we promise, what we never do, and where to look for the legal detail.
01 / 06 WHAT WE PROMISE
Not to advertisers, not to data brokers, not to anyone. This isn't a business model we'll discover later. It's a hard line.
AI outputs are decision support. Educators may review, override, or adjust any AI-generated result. Learners and parents may appeal.
Learner voice, and transcripts are not used to train better-ed's models or any third-party foundation model we rely on. What happens in your classroom stays for your classroom.
Teachers and school administrators control what is assessed, when identification is used, and what is retained. We act on the school's instruction, not the other way around.
We use voice to assess understanding. Teachers control when student identification is used. Voice data is encrypted, access-controlled and never leaves the assessment pipeline.
Voice and assessment data is processed and stored in the region where your school operates, and is not replicated outside that region for ordinary processing.
02 / 06 COMPLIANCE
Better-ed is built to meet the privacy laws of every market it operates in. These laws shape the substance of the product, not just the wording on this page. The frameworks below show how that works in the regions we serve today.
Designed in. EU learner data does not leave the EEA for ordinary processing. Cross-border access by operational support staff is governed by Module 3 Standard Contractual Clauses and a documented Transfer Impact Assessment. Article 27 representative appointed. Article 22 appeal mechanism in place. 72-hour breach notification.
Aligned with NPC Advisory 2024-04 on AI use, including bias-monitoring obligations. Aligned with DepEd Order No. 003, s. 2026, including parental consent for all learners under 18, no learner-facing UI for K to Grade 3, and supervised use for Grades 4 to 12.
Aligned with federal protections including COPPA (children under 13) and FERPA, under which better-ed acts as a "school official" subject to the institution's direction. State requirements include California SOPIPA, New York Education Law 2-d, and Illinois SOPPA.
Educational AI is classified as high-risk under Annex III. better-ed is designed against the high-risk obligations: documented risk management, data governance for training and evaluation data, transparency to deployers and learners, meaningful human oversight by teachers, and accuracy and robustness testing.
better-ed's compliance posture is operational, not aspirational. As we enter new markets, we map local privacy law to the strictest applicable standard already in the product and document the gaps before launch. SOC 2 Type II and ISO 27001 are in progress and will be listed here once the reports are issued. We do not claim compliance with a law, or a certification, we have not operationalized. If your procurement process needs them sooner, existing control documentation can be shared under NDA.
03 / 06 AI GOVERNANCE
Better-ed classifies AI-generated assessment outputs as high-risk under both the EU AI Act risk framework (as imported by DepEd Order No. 003, s. 2026, Principle 4) and DepEd Order No. 003 Annex B. The controls below follow from that classification.
Educators retain final authority on any AI-generated result.
Users see when AI is in use, how it works, and what its limitations are.
Decision logs are retained sufficient to support DPIAs, supervisory enquiries, and the appeal route below.
Aligned with EU AI Act risk-management obligations and NPC Advisory 2024-04.
Where an AI output is used as input to a decision with legal or similarly significant effects for a pupil, the pupil (or parent/guardian) can obtain human intervention, express a point of view, and contest the decision. The route is operated by the school; better-ed supplies the decision logs and explanatory information.
04 / 06 FOR PROCUREMENT TEAMS
better-ed signs the standard Data Processing Agreement for institutional deployments. The DPA incorporates Commission Implementing Decision (EU) 2021/915 (controller-to-processor Article 28 clauses) and, for the sub-processing relationship between Better Oy and Predictive Systems, Inc., Commission Implementing Decision (EU) 2021/914 Module 3 (processor-to-processor SCCs). A Transfer Impact Assessment accompanies the SCCs.
Request the DPA, TIA, sub-processor list with regions, and existing security control documentation at privacy@better-ed.ai. If an NDA is required first, send it and better-ed will countersign.
05 / 06 FOR PARENTS
better-ed lets a pupil answer questions out loud — like a regular classroom recitation, but with a tool that helps the teacher hear and review responses. The AI listens to what was said and gives the teacher a draft assessment. The teacher decides the final mark.
The teachers and administrators at your child's school who are assigned to that subject. better-ed staff only access pupil voice data under documented support requests, with logging. Recordings are never shared for advertising or with third parties beyond the operational sub-processors listed in the Terms of Service.
No. AI produces a draft, and the teacher reviews and decides. If a decision is consequential (a grade that affects a school record), you can ask the school to invoke the appeal route described above.
Yes. For school-deployed use, the school is the controller — start with the school's data-protection contact. better-ed supports the school in handling the request. For Philippines homeschool sign-ups (when available), email privacy@better-ed.ai directly.
Pupil voice recordings default to deletion 90 days after the end of the academic year unless the school has instructed otherwise. Transcripts and assessment outputs are retained for the contract duration plus a 30-day return/deletion window. The full retention table is in the Privacy Policy.
06 / 06 CONTACT
