Responsible AI
Responsible AI policy
The commitments that will apply to every AI feature of Momin Intelligence from the day it is offered to you.
Where things stand today
The Ask Momin page on this site today is not an AI model. It looks up answers in a fixed, approved knowledge base inside your browser, cites its sources, sends your questions nowhere and cannot take any action. It is labelled as automated so that you never mistake it for a person.
Momin Intelligence, the planned AI layer, will answer from published material, help you draft and summarise, search what you are allowed to see and, when you ask and approve, prepare actions. No AI capability will be described as available until it has passed the checks in "Before any AI goes live".
Our commitments
| Commitment | What it means for you |
|---|---|
| You will always know when you are dealing with AI | AI features and AI-generated content are clearly labelled |
| AI acts only within your permissions | It sees and does only what you could see and do yourself |
| People approve consequential actions | Before anything is sent, booked, paid or changed, a person confirms it |
| No training without agreement | Your data is never used to train third-party AI models without your explicit agreement |
| No autonomous decisions about people | AI never decides on employment, credit, education or legal matters |
| You can switch it off | You can turn off AI personalisation; a business can turn AI off for its organisation |
A person stays in charge
AI prepares and suggests; people decide. Every AI action is graded by how much it could affect you or others, and the grade decides how much human involvement it needs.
| What the AI does | What happens before it takes effect |
|---|---|
| Reads and answers | Nothing changes; the answer shows its sources |
| Drafts something you can edit | Nothing happens until you choose to use it |
| Prepares a routine action, such as a message or a booking | You see a preview and confirm it |
| Prepares a sensitive action | A named approver other than you decides |
| Prepares a critical action, such as anything involving money | Two people approve, or the AI may not do it at all |
What AI never decides
Some decisions are never taken by AI, with or without approval. AI may help a person gather information and draft, but a person takes and records the decision.
- Hiring, dismissal, promotion or pay.
- Credit or eligibility for finance.
- Anyone's legal position, legal advice or a legal commitment.
- Payments, refunds, transfers or changes to financial records.
- Approval or signature of a contract.
- Deletion of a customer, an organisation or their data.
- Changes to roles, permissions or credentials.
Our AI does not give legal, financial or investment advice.
Your data and AI
AI will act as a separate identity with the smallest set of permissions it needs, never with your password. Content is filtered before it reaches a model, so the AI sees only what you are allowed to see, and information from one organisation is never used to answer a question from another.
Before anything is sent to an AI model, personal and confidential details the task does not need are removed. Model providers act as our processors under contracts that forbid them from training on your data unless you have agreed. Your data is not training data.
Accuracy and limits
AI can be wrong, and it can sound confident when it is. Our AI is designed to show its sources, to say how certain it is, and to say "I don't have evidence for that" rather than guess. Every AI output is a draft for you to check, not advice to rely on.
Before any AI goes live
Each AI capability must pass these steps in order. A feature cannot skip one, and passing a step for one feature does not pass it for another.
| Step | What must be true |
|---|---|
| Provider | A provider is chosen and contracted, with data processing, region and retention terms |
| Internal use | Audit, budgets and off-switches work; a baseline evaluation has been run |
| First offered to users | AI notices and labels; a data protection impact assessment; permission-filtered retrieval; safety testing; this policy published |
| Actions | Previews and approval flows; each agent certified |
Every AI feature can be switched off quickly, and the rest of the service keeps working without it.
Questions, complaints and review by a person
If an automated or AI answer is harmful, wrong in a way that matters, or shows you something you should not have seen, tell us through the contact page. If an outcome that affects you was assisted by AI, you can ask for it to be reviewed by a person. We tell you what we found and what we did. Your rights over your personal data are explained in the privacy notice.
One standard everywhere
These commitments apply in full in Pakistan, Oman and the European Union. Where a country's law requires more, we meet it and publish the addition here; we never apply a lower standard where the law requires less. If a change to this policy would reduce a commitment, we tell you before it takes effect.