AI Use Notice

What this notice covers

This notice is part of our Terms of Service and Privacy Policy. It applies to every feature that uses a large language model or similar AI system (an “AI feature”), and tells you what each one does, what information it receives, and what control you have. The AI features are:

• Shams Agent, an assistant for school staff, parents and (where enabled) students, that looks things up, drafts messages and documents and, only with the person’s approval, does work in Shams AI on their behalf.
• The learning coach, a chat helper for students about their own Quran work, available only with a parent’s permission for children.
• AI-written reports and feedback for a student, a class, a programme or a school, including weekly summary emails to parents and teachers and scheduled reports.
• Record auto-fill (turning a teacher’s typed description of a lesson into form fields), Hifz Planner explanations (words around a forecast the software calculates), class and programme analysis text, alert wording (optional), book-lesson suggestions from an uploaded book, and column matching when you import a spreadsheet of students.

The old “Hidaya” chat assistant is retired and does not run.

When AI is used on you or your data

• Only when it is switched on for you. A school or independent teacher decides which AI features are available, to which roles, and how far they may act. Schools can switch AI features off entirely.
• Only with consent. Before you use an AI feature we ask you to agree to the use of AI with your information. An adult may switch Shams AI off for their account at any time in Settings, which also pauses scheduled AI reports they own.
• Children need a parent. Any AI feature that a person under 18 uses themselves needs the permission of a parent or guardian, recorded against that child. A child must also be at or above our minimum age for the learning coach (currently 8). Students who are 18 or older decide for themselves.
• AI is not required to use the core of the Service. Record keeping, attendance, progress tracking and messaging work without AI.

What is sent to an AI service

An AI feature sends to an AI provider only the text it needs for that task: your question or instruction, the relevant records it was authorised to read, and fixed instructions that tell the AI how to behave. We reduce what is sent in these ways:

• Placeholders instead of names. For Shams Agent, reports, feedback and weekly summaries, record auto-fill, and planning explanations, our systems replace the names of students, parents, teachers and schools with placeholders (for example “[[student-1]]”), remove email addresses, phone numbers, links and identifiers, and turn a date of birth into an age before the text leaves our systems. The AI’s answer comes back with placeholders, and our systems put the real names back for you. A safety check blocks a request if it still contains a name our systems replaced, and refuses rather than sending it.
• Shapes instead of values. When you import a spreadsheet, only the pattern of each cell (for example “Aaa Aaaa” for a name or “99/99/9999” for a date) is sent so the AI can tell which column is which, never the names, dates, emails or phone numbers themselves.
• Pooled figures. Class and programme analysis text is written from combined statistics and names no student.
• Free text is treated as untrusted. Comments and notes written by people are passed as data, never as instructions, and names inside them are replaced.
• The learning coach is different. To talk to a child naturally, the coach receives the child’s first name, an age band, the date, what the child types, and the child’s own lessons, progress and badges that it reads for that child alone (nothing about other students). It does not receive the child’s surname, date of birth, school name, contact details or ids, and it is told never to ask the child for personal information.
• Book lessons. When a school or teacher asks AI to find the lessons in an uploaded book, the book’s title and the text of its PDF are sent. Upload only books you may share with an AI service.
• Our limits. Replacing names cannot be perfect: unusual spellings, nicknames, or details that identify a person indirectly (such as a rare combination of facts) may still appear in free text. Do not type sensitive information into an AI feature unless needed. If a name you type is ambiguous, the Agent asks which person you mean instead of guessing.

AI providers are listed in the Privacy Policy. One provider is used at a time, with an optional fallback provider if the first is unavailable; the provider currently in use is shown in the assistant panel. Providers receive text through a business or API service, process it to produce the answer and may keep it for a limited time for security or abuse monitoring under their own terms. They are located in various countries, so information may cross borders as described in the Privacy Policy. We do not use the content for advertising and we do not use it to train our own models. We choose providers and settings with the aim that the content is not used to train the provider’s general models, but providers’ practices differ and are governed by their own terms, which we cannot control.

What the Shams Agent can and cannot do

• It acts as you, not above you. It can do only what your own role and permissions already allow, and it checks your access each time. It cannot see a student you cannot see.
• You approve changes. It can read and look things up without asking (you can choose “Ask before reading”), but anything that changes information or contacts a person waits for your explicit approval, shown to you in plain words before it happens. Typing “yes” in chat does not approve anything; you must use the approval button. Changes can be undone within a limited time (currently 30 minutes) where possible.
• It writes drafts, you send. A message to a family or colleague is a draft until you approve it, and each person’s message shows only that person’s own information.
• It cannot change sign-in details, passwords or billing, make payments, delete accounts, handle privacy requests, or perform actions outside the list we have approved for it. It never writes or corrects Qur’ān text.
• It keeps a log. Each run, step, approval and result is logged in a plain summary so that a school can see what the agent did.

Religious content

AI features are instructed to speak within mainstream Sunni scholarship, to say plainly where recognised scholars differ, and to refer questions about rulings, personal circumstances, a student’s own recitation or tajweed to the student’s teacher or a qualified scholar. They do not issue fatāwā, do not judge anyone’s faith, do not cite hadith unless it is supplied from a reviewed reference list, and do not write, complete, correct or translate Qur’ān text; they refer to passages by surah and āyah numbers. They can still make mistakes. AI output is never a religious ruling or a substitute for a teacher.

Accuracy, bias and human review

• AI output can be inaccurate, incomplete, outdated, biased or inappropriate, and can present guesses as facts. Numbers in reports come from our software’s calculations; the AI explains them, and checks reject some answers that do not match the data, but no check is perfect.
• You must review an AI report, summary, draft or suggestion before you rely on it, share it or send it. A teacher, not the AI, makes educational judgements, marks work, assigns grades and decides on discipline or welfare matters.
• We do not use AI to make a decision with legal or similarly significant effect on a person (such as admission, exclusion, grading or pay) without meaningful human involvement. You must not use AI output that way either.
• AI-generated content is identified as such where practical, and AI features tell you they are AI.
• Alerts raised by the Service (such as a drop in attendance) are prompts for a person to look, not conclusions about a child.

Children and the learning coach

• The coach is not a teacher, a counsellor, a doctor or an emergency service. If a child is in danger or distress, they should tell a trusted adult or call local emergency services.
• It does not mark work, contact anyone or change records, and it answers only about the child’s own Quran work and how to use Shams AI.
• Parents can read their child’s coach conversations where the school allows it, can ask for them to be deleted, and can turn the coach off at any time; a school can switch it off for everyone. A child cannot delete their own coach conversations.
• A daily message limit applies to each child. Responses are checked before a child sees them, and requests about religious rulings receive a fixed referral to the teacher, with no AI call.

Records we keep about AI use

Each person has an hourly prompt limit set by the school and by us, to protect against misuse and cost. Automated runs (such as a weekly summary a school has scheduled) are recorded but are not counted as a person’s prompts.

Your choices and rights

• Switch Shams AI off or on for your account in Settings; withdraw consent at any time.
• Parents can grant, refuse or withdraw a child’s AI permission, and read or ask to delete a child’s coach conversations.
• Schools can turn individual AI features off, limit them to some roles, set how far the Agent may act, and view how much AI is used.
• You can ask us for a copy of your AI conversations, ask us to correct or delete them, or object to AI processing by writing to [email protected]. Your rights are explained in the Privacy Policy.

Responsibilities of schools, teachers and other users

If you use an AI feature in your role, you must: do so only for legitimate educational or administrative purposes; have the authority and any consent needed to process the information you give it; not enter more personal or sensitive information than is needed; review the output; and not try to defeat protections such as placeholders, approvals and limits. Schools are responsible for deciding whether, for whom and how AI features are used in their workspace, for telling parents and students, and for training staff.

Changes and contact

AI technology and the law change quickly. We may add, change or remove AI features and providers, and we will update this notice and, for material changes that affect your information, tell you as the Privacy Policy describes. Questions and requests: [email protected].