AI Transparency & System Disclosure
Quick Summary (TL;DR)
- MirrorBuddy uses artificial intelligence to personalize teaching and support inclusive learning
- All significant AI decisions are subject to human oversight per the EU AI Act
- You have the right to request an explanation of how AI has affected your educational experience
- Your data is protected and never used to discriminate or profile without your consent
- You can disable personalized AI at any time and continue using the app with generic settings
11. AI System Disclosure
In compliance with Article 50 of the EU AI Act (Regulation 2024/1689), MirrorBuddy discloses that the 26 educational maestri available on the platform use generative artificial intelligence. These AI systems are designed to personalize teaching and support students with learning differences.
Importante: The AI systems in MirrorBuddy include: Educational Personalization Systems (Azure OpenAI GPT), Progress Analysis Systems (FSRS algorithms for spaced repetition), and Emotional Support Systems (personalized feedback and encouragement generation).
22. AI Classification and Regulatory Compliance
MirrorBuddy utilizza Azure OpenAI (GPT-4), Azure Openai Gpt4, The EU AI Act classifies AI systems by risk level. MirrorBuddy has conducted a Data Protection Impact Assessment (DPIA) and AI Risk Assessment to determine the required compliance level.
MirrorBuddy's Risk Classification
- Limited Risk - Personalization Systems: Educational personalization systems (maestro recommendations, difficulty level adjustments) are classified as limited risk. We use transparent warnings and allow parents to monitor AI decisions.
- High Risk - Where Applicable: If MirrorBuddy integrates automated assessment features in the future that could significantly affect the student's educational path (Art. 6, AI Act), we will implement high-risk controls including mandatory human oversight.
- Compliance with Italian Law 132/2025: MirrorBuddy, in compliance with Law 132/2025 (national implementation of the EU AI Act), maintains publicly available compliance records subject to regular review by the Data Protection Authority.
33. Main AI System Functions
The following AI systems are used in MirrorBuddy to support inclusive teaching:
Described AI Systems
- Educational Personalization Engine: Conversational data (messages with maestri), analyzed in real time to personalize responses and tone
- FSRS Progress Tracking System: Historical performance on flashcards and quizzes, used to calculate optimal review intervals
- Encouragement Message Generator: Emotional context of the study session, used to generate personalized motivational messages
- Learning Difficulty Detector: Recurring difficulty patterns, monitored to send alerts to educators without automatic student categorization
- Semantic Document Search: Vector embeddings of uploaded documents, used for semantic search without storing original content in vectors
All these systems are designed with integrated human oversight and can be disabled in privacy settings.
44. Human Oversight and Control (Article 14 - AI Act)
Human Oversight Controls Implemented
- Parental Oversight: Parents have access to a Parent Dashboard showing all AI decisions about their child, including maestro recommendations and suggested learning paths.
- Educator Oversight: Teachers and school coordinators can view and manually modify AI system decisions. No AI decision is final without human approval.
- Mandatory Human Intervention: If the system identifies significant difficulties, a human educator must intervene within 7 days to provide personalized support. AI provides suggestions only, not decisions.
- Fairness and Transparency Audits: We conduct quarterly audits to verify that AI recommendations do not discriminate against students on the basis of gender, ethnicity, disability or socioeconomic status.
- AI Disabling - Right to Opt Out: Students and parents can completely disable AI personalization at any time. MirrorBuddy still provides all core educational features with generic settings.
Our AI Philosophy
Inspired by AI Safety Research
MirrorBuddy's safety approach is informed by contemporary AI"The Adolescence of Technology" (January 2026), which highlights unique risks of personalized AI in
We consciously choose NOT to be an AI that shapes opinions or createsdependency. Our Professors are designed to be tools of empowerment,
The Professors' Constitution
Our 32 AI Professors operate under a philosophical constitution with sixcore articles:
1. Autonomy First
Every interaction leaves the student MORE capable of facing the nextchallenge on their own.
2. Human Relationships Are Irreplaceable
Professors do not compete with parents, teachers, or friends. They
3. No Opinions, Only Knowledge
Facts and perspectives, never personal opinions on political,
4. Protection from Dependency
Excessive use is a problem, not success. We monitor patterns and
5. Responsible Knowledge
Science education yes, harm instructions no. Some knowledge requires
6. Total Transparency
Students and parents always know they are interacting with AI.Limits and policies are public.
"MirrorBuddy: We amplify human potential, we don't replace
55. Your Rights Regarding AI
The EU AI Act and GDPR grant you specific rights regarding AI systems. Here is what you can do:
- Right of Access and Explanation: You can request to know how the AI made a decision concerning you, including the logic used and the factors considered. Contact roberdan@fightthestroke.org to exercise this right.
- Right to Data Portability: You can export all data used by our AI decisions (conversations, progress, preferences) in a readable format (JSON, CSV) to transfer to another platform.
- Right to Object: You can object to data processing for AI personalization and educational profiling. MirrorBuddy will immediately disable all AI systems for your account.
- Right to Rectification: If you believe the data used by the AI is inaccurate (e.g. the assigned difficulty level), you can request that it be corrected by resetting the AI profiles with accurate data.
- Right to be Forgotten - Erasure: You can request that all your AI data be deleted. This will include the personalization history, while keeping core educational features functional.
- Right to Appeal: If you believe an AI decision has illegally discriminated against you, you can file a complaint with the Data Protection Authority (www.garanteprivacy.it) within 30 days.
66. Data Protection and Security
We implement specific security measures to protect how AI uses your data:
- Data Encryption: Data used to personalize the AI (chat history, progress, preferences) is encrypted end-to-end during transmission to Azure OpenAI.
- Model Anonymization: MirrorBuddy's generative AI models are never trained on students' personal data. We only use real-time input to adapt responses.
- No Discriminatory Profiling: AI does not create profiles that categorize students by cognitive abilities, disabilities, ethnicity or socioeconomic status in ways that could affect educational opportunities.
- Limited Human Access: Only our Italian team of security engineers and data protection officer can view data used by AI for audit purposes. Access is logged and subject to review.
- Limited Retention: AI personalization data is retained only during the student's active session. Historical data is aggregated and pseudonymized within 90 days.
77. AI Risk Analysis and Mitigation Measures
MirrorBuddy has identified the following potential AI risks and implemented mitigation measures in compliance with Article 29 of the EU AI Act:
| Identified Risk | Mitigation Measure |
|---|---|
| Bias in Personalization Algorithms | Bimonthly fairness audits using diverse datasets. Testing on students of various genders, ethnicities and disabilities to verify unbiased recommendations. |
| Over-Reliance on AI | MirrorBuddy encourages human interaction and manual control. The maestri are the primary interface. AI provides only non-binding suggestions. |
| Unauthorized Disclosure of Sensitive Data | End-to-end encryption, data segregation, role-based access policies, permanent audit logs, no extra-EU transfers without consent. |
| Generation of Harmful Content | Safety filters built into Azure OpenAI, human review of generated content, feedback loops for reporting issues, instant disabling if detected. |
| Lack of Transparency in AI Decisions | This document provides full disclosure. Parent Dashboard shows AI reasoning. Right to explanation guaranteed on request. |
88. Regulatory Compliance and Oversight
The Data Protection Authority has the right to inspect MirrorBuddy at any time to verify compliance.
- EU AI Act (Regulation 2024/1689) - Risk classification and transparency obligations for AI systems in education
- GDPR (EU Regulation 2016/679) - Personal data protection, right to explanation of automated decisions (Art. 22)
- Law 132/2025 (Italy) - National implementation of the EU AI Act with specific requirements for education
- WCAG 2.1 Level AA - Accessibility standards for users with disabilities, compliant with the Stanca Law (L. 4/2004)
MirrorBuddy complies with all requirements of the EU AI Act 2024/1689 and Italian Law 132/2025.
99. Contact Us - Questions about AI Transparency
Have questions about our use of AI? You can contact us in the following ways:
- AI Compliance Email roberdan@fightthestroke.org
- Report Form Use the contact form at www.mirrorbuddy.com/contact to send us feedback on AI transparency. We respond to all requests within 5 business days.
- Supervisory Authority Data Protection Authority (garante@gpdp.it)
We are continuously improving AI transparency. Your feedback helps us build a more responsible and inclusive educational platform.
1010. Frequently Asked Questions about AI Transparency
How can I find out why the AI recommended a particular maestro?
When AI recommends a maestro, the Parent Dashboard shows the reasoning (e.g. 'Recommended based on maths performance'). You can contact roberdan@fightthestroke.org for a detailed explanation.
Does my disability affect AI recommendations?
No. AI does not use disability as a direct recommendation factor. We only use the student's learning profile (speed, preferred styles) for universal inclusive teaching.
Can I delete my AI data without deleting my account?
Yes. You can disable AI personalization in Privacy settings. All personalization data will be deleted, but your account and educational content will remain intact.
Which AI models are you using?
MirrorBuddy uses Azure OpenAI (GPT-4o mini for Trial, GPT-5.2-edu for Base, GPT-5.2-chat for Pro) and proprietary FSRS algorithms for memory tracking. No model is trained on MirrorBuddy student data.
What happens if you discover the AI is discriminating against students?
If we detect discrimination (bias), the system is disabled immediately. Data undergoes human audit. Students and parents are notified. If the discrimination caused educational harm, we offer priority corrective support.
11. Compliance Information and Review
This AI transparency document has been prepared in compliance with Article 50 of the EU AI Act (Regulation 2024/1689) and Italian Law 132/2025. It is publicly available in Italian, English, French, German and Spanish.
Next Review: 27 July 2026 (quarterly)