Voice, inbound and outbound
Recordings and transcripts on both directions, whether the agent answered the line or placed the call.
Deze pagina is voorlopig alleen in het Engels beschikbaar. Naar de Engelse pagina
Voice Anonymizer
Masked at source. Before it is ever stored.
Voice Anonymizer finds personal data in a call as it happens and masks it in the copy that gets kept, card numbers, national ID numbers, phone numbers, addresses, names on request, so the stored recording, transcript, report and export never carry it in the clear. The agent still has what it needs to finish the task; masking applies to what is retained, not to what the agent hears.
Masked at source
The task completes. The record does not remember the number.
During the live call the agent works with the data it is given, a card number for a deposit, an address for delivery, whatever the task needs. Masking happens on the way into storage: by the time a recording or transcript is written, the personal data in it is replaced with a masked token. That replacement is irreversible in the stored copy, there is no stored original to reveal later.
- Card and account numbers masked to the last digits
- National ID numbers masked as configured per tenant
- Phone numbers and addresses replaced with a masked token
- Names masked on request, per tenant policy
Caller saidCard number 4532 1191 0000 4821, for the deposit.
Stored asCard number **** **** **** 4821, for the deposit.
Caller saidSend it to 14 Bağdat Caddesi, Kadıköy, Istanbul.
Stored asSend it to the masked address on file.
Voice de-identification
An optional layer for QA and model evaluation.
Beyond masking what was said, Voice Anonymizer can transform the speaker's voice itself in the recordings you retain, so a reviewer or an evaluation set cannot identify the caller by voice. This is a separate, configurable mode, set per tenant, typically applied to recordings kept for quality review or for building evaluation sets, not to every stored call by default.
Where it applies
One rule, everywhere the conversation goes.
Masking is not a setting on one channel. It follows the conversation and its record wherever both are kept.
WhatsApp and chat
Transcripts on messaging channels are masked the same way, so a photo caption or a typed reply carrying a number is treated no differently than a spoken one.
Monthly reports and dashboards
Masked fields appear as tokens wherever they are surfaced: the monthly report, the per-agent dashboard, any view your team opens.
Exports
CSV exports, webhooks and any downstream delivery carry the masked version. What leaves Dring is what was kept, not what was said.
Controls
Configured per tenant, as defined in your contract.
Which categories are masked, how long anything is kept, and who can ever see an unmasked field, all of it is a setting, not a default applied to everyone the same way.
Category list per tenant
Card and account numbers, national IDs, phone numbers, addresses and names are configured independently, on or off per tenant.
Retention per category
Retention windows can differ by category, consistent with the data minimisation principle in KVKK and GDPR: keep only what the task needs, for as long as it is needed.
Audit log
If your policy allows an unmasked field to be viewed at all, that access is written to an audit log: who, when, which field.
Region-locked storage
Recordings, transcripts and masked fields are stored in the region your contract requires. See security and compliance.
Every setting is per tenant and visible in the admin panel: which categories are masked, how long each is retained, and who can request an unmasked view. Nothing here is a platform-wide default applied the same way to every customer.
Categories oncard numbers, national IDs, phone numbers, addresses
Voice de-identificationon for QA set
Retentioncards 30 days, IDs 90 days, addresses 12 months (example)
Regionas defined in contract
How it fits
Part of the stack, not a bolt-on.
Voice Anonymizer runs alongside the products it protects, without changing what they do.
Dring Connect
The summary written back to your CRM carries the masked version of any field that was configured for masking, so the CRM record follows the same rule as the transcript.
Call analytics
Outcome, sentiment and scoring run on the conversation as it happened. Masking changes what is stored, not what is analysed.
Telephony
Masking applies at the edge, where the call is recorded, before the recording is written to storage.
FAQ
Voice Anonymizer questions.
Does masking stop the agent from completing the task on the call?+
No. Masking applies to the copy that is kept, the recording, the transcript, the report. During the live call the agent still works with what the caller gives it, so a deposit, a delivery address or an ID check can still be completed.
Which categories are masked, and can we choose?+
Card and account numbers, national ID numbers, phone numbers and addresses are the standard categories, with names masked on request. The category list is set per tenant, as defined in your contract.
What does voice de-identification do, and is it on by default?+
It is an optional mode that transforms the speaker's voice in retained recordings so the caller cannot be identified by voice, used for QA review and for building evaluation sets. It is configured per tenant rather than applied to every call by default.
Where is masked data stored, and who can see an unmasked field?+
Storage is region-locked to what your contract requires, described on the security page. If your policy permits viewing an unmasked field at all, that access is written to an audit log.
Protect the people on the line.
A short call with an engineer to walk through the category list, retention and where masking fits in your contract.