Security and GDPR
Safe AI Use at School
AI-School is designed for responsible AI use in education with:
- School login capability
- Teacher guidance and monitoring
- Technical and organizational measures for privacy protection
- Option for anonymous accounts without personal data
How AI-School Supports GDPR Compliance
The GDPR requires educational institutions to process personal data carefully and protect the privacy of students. AI-School supports this by:
- Concluding a data processing agreement with schools, which sets out the processing of personal data.
- Using the model agreement from the Education Privacy Covenant as the basis for the data processing agreement.
- Offering prompts anonymously to AI models via a general account, without linking to an identifiable user.
- Informing schools, teachers, and students that no personal data should be included in prompts or documents.
- Using reserved tags and local replacement in document templates, so known data such as school name, address, account name, email address, and role do not need to be sent to the language model.
- Concluding data processing agreements or Data Processing Addenda with the providers of the AI models used.
The school remains the data controller responsible for the chosen setup, authorizations, retention periods, and use within its own organization.
Terminology Framework for Data Processing
Account and student administration data: data such as name, email address, username, and student number that AI-School uses for access, management, and guidance. This data is not automatically added to an AI model.
Prompt content: the question or task a user enters. The chosen AI model receives this content to generate a response. If a user includes personal data in the prompt, the model provider may process that data.
Chat context: previous questions and answers needed to provide a coherent response. Relevant chat context may be sent together with the new prompt to the chosen model.
Additional context: in features such as chatting with documents, images, and tools, selected document fragments, images, search results, or tool results may be added to the model input. This content may also contain personal data.
Subprocessor: an external party that processes data on behalf of the controller, for example for hosting, authentication, or running an AI model. The data processing agreement and annexes govern the current parties, data categories, locations, and safeguards.
Document Templates and Privacy-Sensitive Data
When creating documents via Word templates, AI-School uses a system of reserved tags. For example, the template may contain {school_name}, {school_address}, {account_name}, or {account_email}. The language model then only receives the field name and the instruction to fill the form. The actual personal and organizational data are replaced later by AI-School itself in the application.
This means these known data do not need to be included in the prompt sent to the AI model. This reduces the amount of privacy-sensitive information processed outside the application.
This tag and replacement mechanism is not a substitute for PII protection. It mainly prevents known data from the application from being unnecessarily sent to the language model. If a user enters personal data in a prompt or document themselves, PII protection remains important.
Supervision and Content Filters
Chat history: Teachers have access only to the chat history of students for whom they are responsible. This access can be further limited per role.
Lesson monitoring: Teachers can start lessons and monitor students during the lesson if desired.
Content filters: AI-School and the model providers use safety measures and content filters to limit unwanted or risky output. Such filters reduce risks but cannot guarantee that every unwanted response is blocked.
You can read more about the possible supervision settings under "History Access."
Logging in with School Account
We connect AI-School with Microsoft or Google to enable secure login with your school account. The school must give permission for the use of the Microsoft 365 or Google Workspace integration of AI-School.
The integration checks whether the email address is registered within the AI-School application. This prevents unregistered email addresses from accessing the school environment. No additional personal data beyond the email address is exchanged during this process.
Investigations and Audits
DPIA: Grip IT Consultancy conducted the DPIA in April 2026. The report was finalized on May 20, 2026, and examines which data AI-School processes, who has access, and how data is supplied and withdrawn. Within that scope, the average privacy risk was assessed as low and no high residual risks were identified. The Data Protection Officer reviewed and approved the report. The Data Protection Officer also reviewed and approved the public copy on these pages. The DPIA is a snapshot and can be requested by schools.
Vulnerability scan: Grip IT Consultancy examined ai-school-pro.web.app remotely and partly automated from April 16 to June 16, 2025. The overall score was 8 (low risk), with some points of attention and recommendations. The scan reflects the situation within the examined scope and period and does not guarantee that no vulnerabilities exist.
Information security risk analysis: the analysis of June 19, 2025, maps threats, existing measures, and residual risks, including the risk that a user includes personal data in a prompt. This analysis is also a snapshot.
