Security & Data handling
Call recordings are among the most sensitive data a contact centre holds. Here is exactly how the platform treats yours.
The short version
- Your data stays in India Recordings and everything derived from them are processed and stored on AWS ap-south-1 (Mumbai).
- Sensitive details are redacted With PII masking on, names, account numbers and ID numbers are removed from transcripts.
- Your calls are not training data Recordings and transcripts are used to deliver your QA output. They are not used to train shared or third-party models.
- Access is narrow by default Every screen is permission-gated, users see only their assigned workspaces, and internally only the analyst team assigned to your account can reach your data.
The formal statements live in our Privacy Policy and Terms & Conditions; this page is the operational view — what you configure, and what it changes.
PII masking
Masking is a per-workspace setting under Configuration. Turn it on and personally identifiable details are redacted from transcripts, so a supervisor reviewing a call sees the behaviour without the customer’s identity.
| Aspect | Detail |
|---|---|
| Scope | Per workspace, so a regulated process can mask while an internal one does not. |
| Applies from | Calls processed after the setting is changed. Already-processed transcripts keep the treatment they were processed under. |
| What’s masked | Customer names, account numbers, ID numbers and comparable identifiers in the transcript. |
| Audio | The recording itself is unchanged — masking applies to the transcript and derived text. |
Decide before you backfill
Because masking applies to calls processed after the change, switching it on for a workspace that already holds a year of transcripts doesn't retrospectively redact them. Set masking the way you want it before a historical import.
Where data lives
- Processing and storage are on AWS ap-south-1 (Mumbai). Recordings do not leave India.
- Data derived from your calls — transcripts, sentiment, emotion, scores, flags, summaries — is held in the same region.
- Your own connected storage stays yours. When you connect S3, Google Drive, Dropbox, Box, GCP or Wasabi, the platform reads from your bucket or folder; the source of truth remains under your control.
Retention & Deletion
- Call recordings and derived transcripts are retained for 15 days from processing.
- Audit reports and analyst notes are retained for 12 months unless you ask for earlier deletion.
- Deletion requests go through the contact form and can be made at any time.
Plan around the window
If your own policy requires longer retention of the audio, keep the recordings in your storage — that copy is unaffected by the platform's window. What you keep in the platform is the QA layer, not the archive.
Access controls
- Permission gating Screens and actions render only for accounts that hold the relevant permission — nothing sensitive is merely hidden behind a disabled button.
- Workspace scoping Users reach only the workspaces they were assigned. A supervisor on collections cannot browse sales calls.
- Agent isolation Agents see only their own calls and their own performance, and only when access is granted. They never see other agents, management screens or billing.
- Analyst access Internally, access to your data is restricted to the analyst team assigned to your account.
Set all of this up in Users, Roles & Access.
Source credentials
| Source | What the platform holds | Good practice |
|---|---|---|
| Google Drive · Dropbox · Box | An OAuth authorisation you granted, scoped to reading the folder you selected. | Revoke from the provider’s connected-apps settings when you stop using the integration. |
| Amazon S3 | No keys at all — access comes from a bucket policy you control plus an event notification. | Scope the policy to the single bucket, and remove it to cut access instantly. |
| GCP Bucket | The service-account credentials file you uploaded. | Create a dedicated service account with read-only access to that bucket, nothing wider. |
| Wasabi | The access key and secret you entered, plus bucket and region. | Issue a dedicated read-only key pair, and rotate it on your normal schedule. |
Sign-in & Sessions
- Credentials are encrypted in the browser before they’re transmitted at sign-in.
- Passwords must satisfy strength rules at sign-up, invite acceptance and reset.
- Remember me persists the session on that browser; leave it off on shared machines.
- Password resets go through an emailed link. Links are single-use and expire.
- Sign out from the sidebar when you’re done on a machine that isn’t yours.
Compliance posture
- We operate under applicable Indian data protection law, including the DPDP Act, 2023.
- The platform produces the evidence trail RBI-style reviews ask for: severity-graded findings, week-on-week trends, analyst notes and agent-level scores covering fair-practice language, script adherence and threatening-language patterns.
- We do not claim HIPAA or PCI certification. If your process requires either, talk to us before you commit rather than assuming coverage.
A hardening checklist
- Turn PII masking on for every workspace handling customer identifiers, before your first bulk import.
- Keep admins to two or three. Everyone else is a user with named workspaces.
- Use read-only, dedicated credentials for every connected source, scoped to one bucket or folder.
- Grant agent access deliberately It's a coaching tool, and off is a legitimate default.
- Export the user and agent lists quarterly and review who still needs access.
- Revoke on the day someone leaves User or invite, plus agent access, in one pass.
- Confirm your retention needs against the 15-day recording window, and keep your own archive if you need longer.