For IT leads · DPOs · counsel · regulators

Built for the DPDP Act. Trained for Indic classrooms.

Every layer of Deshika — the architecture, the consent workflow, the safety classifiers, the audit trail, the open-source commitments — is designed for the Indian regulatory and linguistic reality. We tell you exactly what's compliant, what's audited, and what's still to be proven.

DPDP architecture

DPDP-native data architecture. India-hosted only. Schedule IV-aware. School as Data Fiduciary; Deshika as Data Processor. Master Data Processing Agreement (DPA) executed before activation — the 12-clause checklist is published in our open library. India-residency enforced at the AWS-region level, not at app-config. The audit log proves it.

Student and teacher traffic passes through Deshika's DPDP safety layer before reaching the AI model; data stays in India.

Indic safety stack

Bengali + Hindi + Assamese-tuned safety classifiers. 12-layer audit trail per inference. Cert-In incident-reporting in 6 hours; DPDP 72-hour clock to the Data Protection Board. Both clocks run concurrently. Safety classifiers detect distress + harmful content in the student's working language, not just in English. We open-source the classifier weights after Q2 2027.

Open compliance commitments

Policy templates open-sourced after 25 schools. Indic safety classifier weights open-sourced after Q2 2027. Differential-privacy library released with our data-flywheel pledge launch. The moat we want is brand authorship, not lock-in.

The DPDP timeline you're operating against

You have an 18-month runway. Here's how Deshika uses it.

13 NOV 2025
DPDP Rules gazette-notified
13 NOV 2026
Cross-border negative list activates
13 MAY 2027
Substantive enforcement begins

When a student is in distress, the AI hands off. Always.

Deshika's child-safety protocol has three tiers (T1 — distress signal · T2 — active distress with method or plan · T3 — imminent danger). At every tier the AI breaks character, surfaces a human helpline + the school's counsellor, and notifies the pre-registered trusted adult per consent. The AI never plays therapist.

The Trevor Project — the deepest-experienced organisation in the world on youth crisis lines — explicitly chose not to use AI as the front-line. They use AI to train their human counsellors. That's the data point this whole protocol is built around.

Tele-MANAS

14416
Government 24×7 · ~20 Indian languages · free

Vandrevala Foundation

1860-2662-345
NGO 24×7 · call + WhatsApp + chat · free

iCall (TISS)

9152987821
Mon-Sat 8am-10pm · English + Hindi + Marathi

KIRAN

1800-599-0019
24×7 · 13 Indian languages · disability focus

Posting numbers on a poster is not a partnership. Deshika signs warm-handoff partnerships with at least one helpline before V1 ship.

What we'll open-source

The moat we want is brand authorship, not lock-in.

Q3 2026 · after 25 schools

AI Policy Starter Pack

All 7 policy templates, India-customised, DPDP-aligned. Creative Commons CC-BY-SA.

Q2 2027

Indic safety classifier weights

Bengali, Hindi, Assamese-tuned distress + harmful-content classifiers. Replicable evaluation harness.

With data-flywheel pledge launch

Differential-privacy aggregation library

The library that lets us learn from school-level data without ever exposing student-level data.

Terms

Plain-English commitments · v1.0 · pending counsel review before public launch.

Who you contract with

Deshika Technologies Private Limited, an Indian company. Your contracting party is your school — never the student or parent directly. The school is the Data Fiduciary; we are the Data Processor under DPDP §2(k). Operating on the school’s written instruction is the entire shape of the agreement.

What we promise

India residency (AWS Mumbai). Child-safe AI tutor that hands off to verified helplines on distress. No model training on student data. No third-party sale or sharing. Per-school encryption keys. Audit log retained for the full school contract term and 12 months after.

What you owe

Verifiable Parental Consent at enrolment (we provide the workflow), AI policy adoption (we provide the templates), a non-AI learning alternative for opted-out students (we help you design it).

What happens if we mess up

72-hour breach notification to your DPO. Indemnification capped at the prior 12 months of fees per the master DPA. School can terminate for cause if we lose DPDP-significant compliance certifications during the term.

What this is not

This is a plain-English summary, not the master DPA. The signed Data Processing Agreement attached to your school’s contract is the binding document. Counsel review is in progress — expect light language updates, no posture changes.

Cookies

What this website actually stores on your device · v1.0.

Marketing site (this site)

We don’t use third-party analytics, ad cookies, or trackers. The site does write two small items to your browser’s localStorage for usability: your theme choice (light/dark) and a welcome timestamp. Neither leaves your device.

Assessment tool

The school-AI-readiness assessment runs entirely in your browser. Your quiz responses are sent to Deshika only when you submit the lead form. Until then nothing leaves your device.

The product (classroom app)

The signed-in product uses one HTTP-only session cookie (deshika_session) for authentication. No third-party cookies. No ad networks. No cross-site tracking. Per-school encryption keys mean no shared cookie namespace either.

How to clear everything

Use your browser’s “clear site data for deshika.xyz” option (Chrome: Settings → Privacy → Site settings → Cookies). For account data on the product side, file a Data deletion request via the contact form and we’ll process it within the DPDP 30-day window.

Got a specific DPDP question?

Email our DPO. Replies typically within 1 business day. For school IT leads and counsels who need to scrutinise before recommending to the board.

Email our DPO