Dexter AI Data-Handling Statement
Last updated: 7/7/2026
Dexter includes an in-app assistant (“Dexter”) that uses artificial-intelligence (AI) language models to help students plan their coursework, prepare for assessments, and answer questions about their own courses. Because institutions are appropriately careful about student data and AI, this statement describes exactly what data is sent to AI providers, why, and under what terms.
1. When data is sent to an AI provider
Dexter sends data to an AI provider only when it is needed to generate a response for the student - for example, when a student asks the assistant a question, or when the app produces a study plan or summary. Dexter does not send student data to AI providers for any other purpose, and does not send it on a continuous or background basis unrelated to a student’s use of the assistant.
2. What data is sent (data minimization)
Dexter sends only the context relevant to the request, drawn from the student’s own connected course materials and the student’s question - for example, a relevant excerpt of a syllabus, assignment, or course document, together with the student’s prompt. Dexter does not send a student’s full record, other students’ data, or credentials. Where practical, only the minimum context needed to answer is included.
3. Which AI providers we use
All Dexter AI models are hosted in the United States. Dexter uses the following AI provider(s) to generate responses:
Google
Anthropic
OpenAI
Below is a breakdown of the different models used and how they are used:
Main chat / Q&A — Gemini (gemini-2.5-flash)
Nuanced chat — Gemini (gemini-2.5-flash)
Assignment coach — Gemini (gemini-2.5-flash)
Router / intent classification — Gemini (gemini-2.5-flash-lite)
Draft / PDF extraction — Gemini (gemini-2.5-flash)
Today coaching (“tip of the day”) — Anthropic (claude-sonnet-4-6)
Assignment readiness / help-me-prepare — Anthropic (claude-sonnet-4-6)
Retrieval embeddings (RAG index + every query) — OpenAI (text-embedding-3-small)
Fallback (every route, when the primary errors) — OpenAI (gpt-*)
Conversation safety (the guard model) — Google via Vertex (openai/gpt-oss-120b)
This list is kept current and is available to institutions on request. Dexter will provide notice before adding a provider that would process student data.
4. Retention and model training
Dexter engages its AI provider(s) under business/API terms - not consumer chatbot terms - under which:
the content Dexter sends is not used to train or improve the provider’s models; and
the content is subject to zero or limited retention by the provider (retained only briefly, if at all, for operational and abuse-prevention purposes, then deleted).
Dexter does not use student data to train its own models.
5. No sale, no advertising
Student data sent to AI providers is used only to generate the student’s own response. It is never sold, and it is never used for advertising or profiling unrelated to the Service.
6. Human oversight and limitations
The assistant provides study and planning support. It does not make decisions that produce legal or similarly significant effects about a student, and it does not assign grades or take academic actions. Students remain responsible for their own work, and the assistant’s output should be treated as guidance, not authoritative fact.
7. Security and processing location
Requests to AI providers are made over encrypted connections (TLS 1.2+). Dexter’s own storage and processing occur in the United States (Google Cloud, us-central1); AI-provider processing locations are listed in Section 3.
8. Contact
Questions about Dexter’s use of AI or requests for provider terms can be sent to info@dexterweb.io.