AI Transparency
Last updated: 22 April 2026
1. Overview
Writing, no kidding uses artificial intelligence to assist teachers in reviewing student writing. This page describes which AI features are active, what model powers them, what data is processed, and what rights you have under the EU AI Act (Regulation (EU) 2024/1689) and the General Data Protection Regulation (GDPR).
We are committed to keeping human teachers firmly in control. Every AI output on this platform is a suggestion — teachers review, approve, or override all AI-generated feedback before students ever see it.
2. AI Features in Use
AI Writing Correction Tool
Analyses a student's written submission and generates suggested corrections — grammar, vocabulary, coherence, and style — together with explanatory feedback. The result is shown to the teacher, who reviews it before releasing feedback to the student. Students are informed when feedback was AI-assisted.
AI Rubric Scoring
Proposes a rubric score and rationale for each criterion in a teacher-defined writing rubric. The teacher sees which rubric level was AI-suggested (marked [AI]) and can accept or override any score before publishing results. No score is ever final until a teacher approves it.
Originality Checker
Produces a probabilistic originality indicator for a student submission. This score is an analytical signal only — it is not evidence of academic misconduct and must never be the sole basis for any integrity decision. Teachers are shown this disclaimer alongside the score.
3. AI Model & Provider
| Parameter | Value |
|---|---|
| Model | Google Gemini 2.5 Flash |
| Provider | Google LLC (Firebase Genkit) |
| Data processing region | EU / EEA (Google Cloud, europe-west1) |
| Provider AI Act status | Google is subject to GPAI obligations as a General-Purpose AI model provider. See Google AI Principles. |
| Training data summary | Google publishes information about Gemini's training data at ai.google/responsibility. Writing, no kidding does not train or fine-tune models. |
4. What Data Is Processed by AI
When a teacher triggers an AI feature, the following data is sent to Google Gemini:
- The student's written submission text.
- The assignment prompt and any rubric criteria defined by the teacher.
- No student names, email addresses, or identifiers are included in AI requests.
AI requests are initiated exclusively by the teacher — students' data is never sent to the AI model without a teacher action.
5. Human Oversight & Override
In compliance with EU AI Act Art. 14 (human oversight), every AI output is presented as a draft to the teacher:
- AI corrections can be edited, deleted, or rejected before the teacher publishes feedback to students.
- AI-suggested rubric scores are clearly marked
[AI]. Teachers click a different level to override; no AI score is submitted without an explicit teacher action. - The Originality Checker result is advisory only. It cannot trigger any automated consequence — the teacher decides all outcomes.
6. Risk Classification
Writing, no kidding has carried out a formal risk classification assessment under EU AI Act Art. 6(3) and Art. 6(4). Our conclusion is that none of our AI features qualify as high-risk systems under Annex III §3 (education and vocational training). The full assessment is documented in Compliance/Art6-4-Assessment-AI006-AI007.md (available on request).
Art. 6(4) Derogation — Key Findings
Although features AI-006 (Rubric Scorer) and AI-007 (Student Evaluation Generator) fall within the scope of Annex III §3(b) (AI systems that evaluate learning outcomes), we assess them as not high-risk on the basis of Art. 6(3)(b) and (d):
- AI output is not binding. AI-006 and AI-007 produce suggestions presented to the teacher. No score, grade, or evaluation reaches a student until the teacher has explicitly reviewed and confirmed it.
- Teacher confirmation is architecturally enforced.The “Save rubric” button is blocked until the teacher has clicked a level for every AI-scored criterion. The “Submit Review” button requires an explicit teacher-entered grade — the AI suggestion is shown read-only and cannot be auto-applied.
- No access, progression or certification decisions. The platform does not determine enrolment, academic progression, or certification. Those decisions remain entirely with the teacher and institution.
- The AI improves a previously completed human activity (Art. 6(3)(d)): it assists the teacher in reviewing work they have already assigned and received. It does not initiate or automate the assessment process.
These controls are architectural guarantees, not procedural claims. They cannot be bypassed by the teacher or by the platform without a code change subject to legal review (see our Quality Management System).
We continue to monitor EU AI Act guidance, Commission delegated acts, and AI Office opinions that may affect this classification. Our annual review cadence is documented in Compliance/Quality-Management-System.md.
7. Data Retention
AI usage logs (token counts, model type, timestamp) are retained for up to two years for billing, auditing, and quality monitoring purposes, after which they are automatically deleted via a Firestore TTL policy (EU AI Act Art. 19 / GDPR data minimisation).
Student submission text is retained as long as the teacher account is active or as otherwise specified in our Privacy Policy.
8. Your Rights
Under GDPR and, where applicable, the EU AI Act, you have the right to:
- Access the personal data we hold about you.
- Request correction or erasure of your data.
- Object to or restrict certain processing.
- Receive a meaningful explanation of any AI-assisted output that affects you (Art. 50(1) EU AI Act).
- Not be subject to a decision based solely on automated processing (GDPR Art. 22) — all AI outputs require teacher approval before they affect any student.
To exercise any of these rights, contact us at privacy@writingnokidding.com.
9. Incident Reporting
If you believe an AI feature has caused harm, produced a seriously incorrect output, or violated your rights, please use our AI Incident Reporting form.
10. Contact
Questions about this page or our AI practices can be directed to privacy@writingnokidding.com.