Model Training & Data Usage

Your data runs your system. Nobody else’s. This page is the plain statement of how Anuma Intelligence Pvt. Ltd. uses merchant data with AI, and it is part of our Privacy Policy.

What your data is used for

Your catalog, prices, buyers, conversations, and outcomes are used to operate and improve your Anuma deployment. The system quotes from your rate card, negotiates within your concession ladder, and learns your business’s patterns: which items move together, which buyers reorder, which terms close. That learning is tenant-scoped: it applies to your desk alone and makes your system sharper about your business. Prompts are sent to the model providers named in the Privacy Policy, under agreements that bar them from training on your content.

What we never do

• We do not sell your data. Not to advertisers, not to data brokers, not to anyone.
• We do not use your data to train foundation models. The large language models Anuma orchestrates are provided under agreements that prohibit the provider from training on your content.
• We do not move learning across merchants. Your rates, margins, buyer lists, and negotiation patterns are never shared with, or used to improve outcomes for, any other Anuma merchant. We do not pool merchant data, and we have never run any aggregated processing of it. Small models trained for a single merchant are on the roadmap, and each one would stay that merchant’s own.
• We do not profile people. No behavioural advertising, no third-party marketing, no profiling unrelated to running your orders.

Personal information is redacted before AI processing

Names, phone numbers, and addresses are separated from working data and held in an encrypted vault. What enters a model prompt is the deal: items, quantities, terms, history. Identifiers are restored only at the edge, when a message is actually sent to a person who should receive it.

Every action is logged

Every AI action on your account is recorded with its inputs, the policy version it ran under, and the decision trace: the rule, floor, or approval that produced the outcome. The log is yours to inspect. If the system drafted, said, or sent something, you can see what it acted on.

Aggregate service improvement

We use anonymised, aggregated operational signals (error rates, latency, match confidence, feature usage) to keep the platform reliable and to improve it. This telemetry contains no personal data and no merchant’s commercial content.

When you leave

Your data goes with you: export in a structured, machine-readable format on request, and deletion on the schedule in the Privacy Policy. Anything the system learned from your data is deleted with it.

Questions

Ask us directly: hello@anuma.co.in. If our practice ever changes, this page changes first, and account holders are notified before anything takes effect.