डेटा के नियम
How data is handled, as far as it is true today
This page states only what is in place now. Where something is a plan, it says so.
Three classes of data
Every AI request from an AskShiva platform belongs to one class, and the class decides which AI models may receive it.
| Class | Examples | Which AI models may receive it |
|---|---|---|
| Public | News, public web pages, market data, generated images | Any provider |
| Personal | Chats with an assistant, business records, birth details, HR records, startup ideas, screenshots | Non-China-hosted models only |
| Children's | Anything a school or learning product holds about students | OpenAI models only, with no fallback to other providers, or a model that runs on the phone itself |
How the rules are applied
The platforms call AI through the AskShiva AI Gateway. Each app has its own key with a budget and a list of the model routes it may call, and each route belongs to one data class. A children's product's key lists only the OpenAI routes. An app that handles personal data uses non-China routes for it. A few apps also handle public content, such as news for the AskShiva assistant; their keys also list a public route, and the app uses it only for that public content. When a personal-data route falls back to another model inside the gateway, the fallback is also a non-China provider.
The AI Gateway state is read live when JavaScript is available.
When the gateway does not answer, AI features in some apps are degraded or switched off, and some apps send requests to a provider directly instead. The direct fallbacks recorded in our status data for personal-data work all go to non-China providers. Each platform's status on the platforms page shows its AI state.
Children's products
- Campus One runs a made-up demo school only, and sign-up is disabled. Before any real school, student, parent or teacher data is loaded, it moves to a server in India.
- Vidya AI: the tutor runs on the phone. Its website has no sign-in, no forms and no cookies, and its server part holds demo data only.
- Exam Paper Generator runs inside the Campus One demo with demo questions. AI may help choose among questions in your bank but never writes question text, and a teacher reviews every paper.
- Vidya Copilot is in development; its marking tool stays switched off until it passes our accuracy check.
This website collects no data about anyone, children included.
Where the platforms run today
Our application server and our GPU server are rented servers in the USA. askshiva.ai is still served by a previous server until its move to the new servers. AI features send text to outside model providers under the rules above. The AI Gateway runs on the application server. The GPU server runs the shared image engine and local AI models, and stores no personal data.
Keeping Indian customers' personal data in India is a plan, not yet a fact. Until it is in place, please do not enter sensitive personal information into any AskShiva platform.
What this website does
- Plain pages: no forms, no cookies, no analytics and no trackers.
- No fonts, scripts or images from other sites.
- One kind of extra request: the status files on this same site (
/data/portfolio.jsonand/data/status.json), which hold product status, not visitor data. - Like any web server, ours may record basic technical request data.
What we do not claim
We hold no security or privacy certifications and make no compliance claims. These rules describe how our systems are set up today; they are general information, not legal advice.