Tutorio uses artificial intelligence to help tutors create better lessons faster and to give learners more feedback than any tutor could write by hand. We think you should know exactly where AI is used, what it sees, where it falls short and who is accountable. This statement explains all of that. It forms part of our Terms of Service and Privacy Policy.
Where we use AI
| Feature | What the AI does | Who sees the output first |
|---|---|---|
| Course drafting | Turns a syllabus or specification into a topic structure, drafts explanations and questions, and audits the draft for gaps against the specification | The tutor, in the course editor, before anything is published |
| Marking | Compares a learner's written answer with the mark scheme and suggests a mark and feedback | The learner sees the suggested mark and feedback; the tutor can review and override |
| Feedback and explanations | Explains why an answer is right or wrong and suggests what to revise | The learner |
| Rio | A study companion that answers questions about the topic being studied, offers hints and encouragement, and helps a learner plan revision | The learner |
| Review generation | Picks retrieval-practice questions from material a tutor has already taught, so reviews test only what was covered | The learner, within a tutor-set review |
AI is not used to: decide grades, set predicted grades, place learners in sets, decide who can join a class, make safeguarding decisions, price anything, or target advertising.
Human oversight
- Tutors review and can edit every piece of AI-drafted content before it is published to learners.
- AI marks are suggestions. Tutors can see them, override them and are told when a learner disputes one. Learners can ask their tutor to review any AI-marked answer.
- Screening tests, results and reports remain the tutor's or organisation's decision to interpret and act on.
- We label AI-generated content and feedback so you know what you are looking at.
What data the AI sees
For each request we send only what that request needs:
- For marking: the question, the mark scheme and the learner's answer.
- For Rio: the learner's message, recent turns of the same conversation and the topic and question being studied.
- For course drafting: the syllabus text or checklist the tutor provides and the draft so far.
We do not send learner names, email addresses, ages, class membership or device data with AI requests, and we strip identifiers where we can. If a learner types personal information into Rio it will be sent as part of the message; we advise learners not to do so and Rio is instructed not to ask for it.
Our AI providers
We currently use Anthropic's Claude models through their commercial API for text features, and Google's Gemini API to generate course cover images and icons from a tutor's text prompt (no learner data is involved). Under our agreements the providers may not use the content of our requests to train their models, retain it only briefly for abuse monitoring or not at all where zero-retention is available, and process it under data protection terms. Providers and locations are listed in Sub-processors & Data Locations. If we change provider we will update this statement and notify organisations.
Credits and fair use
AI features consume AI credits. Each plan includes a monthly allowance and organisations can purchase more. Credits protect everyone by keeping usage within limits and by preventing automated abuse. When a learner or tutor runs out of credits the AI feature pauses; nothing else in Tutorio is affected.
Limits you should know about
- AI can be wrong. Language models make confident mistakes, including in maths, dates, quotations and mark-scheme interpretation. Treat AI output as a strong first draft or a second opinion, never as the final word.
- AI cannot see intent. A marked answer reflects what was written, not what the learner meant. Tutors know their learners; AI does not.
- Bias. Models can reflect biases in their training data. We test for unfair marking across writing styles and dialects, and tutors can override any mark.
- Not a person. Rio is a software character. It is designed to be encouraging, but it cannot provide counselling or safeguarding support. If a learner tells Rio something that suggests they are at risk, Rio directs them to a trusted adult and the conversation is flagged for human review under our Safeguarding Policy.
Safety measures
- Rio and other AI features operate under instructions that restrict them to educational content appropriate for children, and provider-side safety filters apply on top.
- Prompts and responses are checked for personal data and inappropriate content.
- Requests are rate-limited and monitored for abuse, including attempts to extract system instructions or other users' data.
- AI conversations are stored with the learner's record so tutors, parents and our safety team can review them when needed.
Your choices
- Learners can choose not to use Rio; nothing in a course requires it.
- Tutors can turn off AI marking for an assignment and mark by hand.
- Organisations can ask us to disable AI features for their whole deployment.
- Anyone can ask us how a specific AI output was produced, or ask for a human to review it, by contacting [email protected].
Governance
We assess new AI features for data protection, fairness and child-safety impact before launch, record the assessment, and review live features at least every six months. We follow the ICO's guidance on AI and data protection and the Department for Education's guidance on generative AI in education.
Changes
We will update this statement whenever we add an AI feature, change provider or materially change how AI is used. Every version is kept in the history below.