Daftar

Privacy Policy

Last updated: 25 May 2026

Daftar ("Daftar", "we", "us") provides AI-powered receipt and invoice extraction for accounting firms across the MENA region. This Privacy Policy explains what data we collect, how we use it, and the rights of accounting firms and their clients whose receipts pass through our platform.

1. Data we process

  • Account data. Name, email address, password hash, preferred language, and the firm you belong to.
  • Receipts and invoices. Documents you (or your clients) upload, email, or capture. These are stored encrypted at rest in Supabase Storage.
  • Extracted fields. Vendor, dates, amounts, taxes, line items, and confidence scores produced by our AI pipeline.
  • Integration data. OAuth tokens for Google (Sheets + Drive) when you connect a destination spreadsheet. Tokens are encrypted at rest.
  • Operational logs. Request metadata, error traces, and SHA256 hashes of receipt files (we never log raw image bytes).
  • Billing data. Plan, usage counts, and Stripe customer / subscription identifiers. Card details are held by Stripe — we never see them.

2. How we use it

  • To extract structured fields from your receipts and present them in the review queue.
  • To sync approved receipts to the Google Sheet you authorise, and to generate CSV / XLSX exports on request.
  • To bill your firm, send transactional email (receipts, alerts, sync failures), and provide support when you ask.
  • To monitor reliability and security — including detecting duplicate uploads and abuse.

3. Sub-processors

We rely on a small set of vetted providers. Each handles only the data needed for their function:

  • Supabase — Postgres database, Auth, and Storage (encrypted at rest).
  • Anthropic (Claude) — primary vision model for extraction. Document images are sent for inference and are not used by Anthropic to train models on our zero-retention configuration.
  • OpenAI / Google (Gemini) — second-pass fallback model when our primary model returns low confidence. Same zero-retention posture applies.
  • Postmark — inbound email ingestion for your client forwarding addresses (*@in.daftar.wannadev.me).
  • Stripe — payment processing and subscription billing.
  • Google — only when you explicitly connect Google Sheets / Drive via OAuth. We request the minimum scopes (spreadsheets and drive.file).
  • Vercel — hosting and edge delivery.
  • Sentry, PostHog — error tracking and product analytics.

4. Tenant isolation

Daftar is multi-tenant by firm. Every database query is scoped to a firm via Postgres Row-Level Security (RLS) — not just application code. One firm cannot read, update, or delete another firm's receipts, clients, exports, forwarding addresses, or extractions. We verify this isolation as part of every release.

5. Retention

  • Receipts and extractions are retained while your subscription is active so your firm can re-export historical books.
  • If you delete a receipt, both the stored file and its extraction row are removed.
  • If your firm cancels and requests deletion, we delete your firm's records within 30 days of confirmation. Backups roll off within 90 days.
  • Operational logs are retained for up to 90 days.

6. AI training

We do not train Daftar's own models on your receipts. Our model providers (Anthropic, OpenAI, Google) are configured for zero data retention where their APIs offer it.

7. Your rights

Depending on your jurisdiction (UAE PDPL, KSA PDPL, Egyptian Personal Data Protection Law, EU GDPR for any EU users) you may have rights to access, correct, export, or delete personal data we hold about you. Email privacy@daftar.wannadev.me and we will respond within 30 days.

8. Security

  • All data is encrypted in transit (TLS 1.2+) and at rest.
  • OAuth tokens are encrypted with envelope encryption at the application layer.
  • Access to production data is limited to the engineers who need it and is audit-logged.

9. Contact

Questions about this policy? Email privacy@daftar.wannadev.me.