This page explains, specifically, how PromptPaddi Junior relates to AI and what happens to a learner's data, separate from our general Privacy Policy. There are two different ways AI appears, and they handle data differently.
1. Study Coach prompts: the learner's own AI tool
Study Coach is a library of ready-made, curriculum-aligned prompts. To use one, a learner copies it and pastes it themselves into an AI tool they already use (for example Claude, ChatGPT, or Gemini) directly on that provider's own site or app. That conversation happens entirely between the learner and that AI provider; it never passes through PromptPaddi Junior's servers, and we do not see, store, or have access to it. Whichever AI tool a learner (or their parent, guardian, or school) chooses is governed by that provider's own terms and privacy policy, not ours.
Teachers who use the free helper in Homework Tracker to draft questions do the same: they copy the request into their own AI account, and no pupil names or answers are included.
Primary pupils can also "Study without AI": Study Coach sessions with a short lesson and questions, marked by our own system. No AI provider is involved, and nothing from them is sent to one.
2. Paddi: our built-in AI tutor
PromptPaddi Tutor, Exam Boost and homework help inside Homework Tracker use Paddi, an AI tutor that runs inside PromptPaddi Junior. Paddi is powered by Anthropic's Claude models, which we reach through Anthropic's API. Schools that add Paddi to Homework Tracker also have Paddi mark written homework answers; the teacher sees each mark and can change it.
What is sent to Anthropic, only when a learner asks Paddi for help, opens a practice set or lesson, or hands in a written answer for Paddi to mark: the subject and topic, the question, the learner's message and the last few messages on that question, whether they have checked their answer and how sure they were, and, for a written answer Paddi marks, the answer itself and the class label (for example "JSS 2 Blue"). Anything a learner types to Paddi is part of what is sent.
What is not sent: the learner's name, their school's name, an email address, or their private link. Please remind learners not to type personal details (their full name, address or phone number) into Paddi.
Scheme-of-work uploads. A parent or form teacher can upload a photo or PDF of a school's scheme of work so Paddi can read the topics out of it for them to check. That file is sent to Anthropic to be read and is not kept by us: we keep only the topics the adult confirms. Please upload only the topic list, not documents that name pupils.
What we keep. We do not store the conversation with Paddi. We keep the learner's answers, marks and scores, how many times Paddi was used each day, and the size and cost of each request (the model used and the amount of text, not the text itself), so the learner, their parents and teachers can see progress and so we can apply daily limits. Anthropic processes what is sent in order to produce the reply, under its API commercial terms and its Data Processing Addendum.
Sending data abroad. Anthropic is based in the United States, so this data leaves Nigeria. Under the Data Processing Addendum Anthropic is our processor, must tell us of a security breach within 48 hours, and must delete data within 30 days of the contract ending. It uses Standard Contractual Clauses for transfers out of Europe and the UK; it does not mention Nigeria, so we treat Nigerian transfers as needing our own safeguards (no names or contact details sent, nothing stored by us, adult supervision).
How long Anthropic keeps it. As of 14 October 2026 Anthropic's published terms say it does not train its models on customer content, and that it deletes API inputs and outputs from its systems within 30 days by default. The exceptions are content flagged under its usage policy (kept for up to two years), anything the law requires it to keep, and any separate agreement with us. Its retention page is at privacy.claude.com.
Safe AI Start. Before Paddi works with a child, they do a one-minute safety check: keep personal details out, check what AI says, and tell a trusted adult if something feels wrong. We keep only that it was done and when, not the answers.
Safeguards. Paddi is instructed to stay on schoolwork, to give hints before answers, to say it is an AI, and to send a learner who mentions feeling unsafe or upset to their form teacher or another trusted adult. When Paddi marks a written answer that suggests a learner may be unsafe or very upset, it does not mark it: the answer goes to the teacher, and the form teacher and the school's named safeguarding contact are emailed. Paddi does not replace a school's safeguarding procedures, and a conversation with Paddi is not monitored by a person.
Learning GPS does not use AI: nothing a pupil writes in the Daily Pulse, Friday Reflection, GPS Check or monthly check-in is sent to an AI provider.
Keeping pupils' names off PromptPaddi Junior
Pupils' names are never sent to an AI provider. A school can go further and switch on "pupils by code": we then hold a code for each pupil instead of their name, and the names stay on the school's own devices. See Pupils by code in our Privacy Policy.
AI can be wrong
Like all AI systems, tools such as Paddi, Claude, ChatGPT, and Gemini can produce answers that sound confident but are inaccurate or outdated, sometimes called "hallucination." PromptPaddi Junior is a study aid, not a source of truth on its own. Our AI Literacy Course teaches learners directly what AI can get wrong and how to spot it. Important answers, especially for exams and coursework, should always be checked against a textbook, teacher, or trusted adult.
Questions
For anything about how PromptPaddi Junior relates to AI or handles a learner's data, email hello@promptpaddijunior.com.