Privacy Policy

Synapse-AI ("we", "us", "our") is a legal-tech assistant for Indian advocates. This policy explains what data we collect, how we use it, where it lives, and what rights you have.

What we collect

Account data: email, full name, optional bar council ID, optional specialization. Captured during signup or filled in later from Settings.
Case data: case titles, descriptions, parties, transcripts, extracted facts, drafts, hearings, and any documents you attach. All case data is owned by the advocate who created the case row.
Audio: opus-encoded recordings from the Quick Consult flow. Audio is uploaded resumably to Supabase Storage in the India-South region. Recordings are retained until you trigger a hard purge or the underlying case is deleted.
Usage telemetry: per-call cost tags (Anthropic / Voyage / Sarvam / IndianKanoon), error rates, and request latency. Used for billing reconciliation, performance triage, and SLO tracking.

Where data lives

Region: Supabase project locked to India-South (Mumbai/ Hyderabad). Sarvam Saaras v3 (ASR) calls also run in the India-South region.
Out-of-region egress: Anthropic Claude does not yet host an India-region inference endpoint. Calls to Claude originate from the India-hosted FastAPI backend; before any text leaves India, party names and other personally-identifiable strings are tokenised via the PII Redactor. The redaction map is held in-memory per session and discarded on session close. Re-hydration of redacted tokens happens on the response path, also inside the India-hosted FastAPI process.
Vendor terms: Anthropic's commercial terms retain inputs for up to 30 days for abuse review. We have a Zero Data Retention contract in negotiation; until it lands, this 30-day retention applies. Anthropic does not use commercial-tier inputs for model training.

What you can do with your data

Access: every action you take is captured in an audit log. Read your own log via Settings → Activity log (calls GET /v1/audit/me).
Hard delete: Settings → Purge all my data wipes every case, transcript, document, draft, hearing, audio segment, and research item you own. Storage objects are wiped simultaneously. The auth row + profile remain so you can sign back in to an empty workspace.
Privacy Mode: Settings → Privacy Mode forces all transcribe / extract / draft / research calls through fake clients with redacted inputs. Use this when working on highly-sensitive matters where even the 30-day Anthropic retention window is unacceptable.

Children

Synapse-AI is not directed at children under 18 and we do not knowingly collect data about them.

Updates

We may revise this policy. The "Last updated" date at the top of this page reflects the current version. Material changes will be surfaced in-app on next sign-in.

Contact

For privacy questions, email privacy@synapse.ai. For DPDPA-specific queries, email dpo@synapse.ai (Data Protection Officer mailbox).