Home → Privacy Policy

Privacy Policy

05.11.2026

1. What data is processed

2. Why

To provide the service, calculate charges and issue invoices, prevent abuse and offer support. The data is not used for advertising or profiling and is never sold.

3. Where processing happens

Recognition runs on the operator's own hardware — audio is not sent to third-party recognition services and is not used to train models. For clients with analysis enabled, a local model on our own hardware processes the transcript, without third-party analysis services. To open a dispute, you provide your score and a description of your disagreement; the local model reviews the assessment, and the original assessment is preserved and shown alongside the final one. The complaint, assessments, decision and a copy of the call transcript are stored for your account only. After an accepted or partially accepted decision, the final assessment and transcript serve as a labeling example for similar new calls in that account only. Complaint text is excluded from examples; pending, failed and rejected disputes are not used as examples. This data is never used to analyze other clients' calls.

Request handling and storage of account data, audio, transcripts and analysis results run on a rented server. Client connections are protected by HTTPS. The Service owner can access the content in the panel and uses it only for technical support and incident investigation.

4. Sharing with third parties

Data is not sold. The payment provider receives what is needed to process a payment, and the hosting provider technically operates the server. Data may be disclosed on a lawful request from a competent authority.

If you enable notifications through your Telegram bot, Telegram receives the risk assessment, category and summary, or the first 200 characters of the transcript when no summary is available. If you configure a webhook, your chosen recipient receives the record identifiers, status and, for completed analysis, the risk score, level and category; the webhook does not include the transcript or summary.

5. Cookies

Only functional cookies are used: the client-area session and the remembered interface language. There are no analytics counters or advertising trackers on the site.

6. Retention and deletion

Audio — 24 hours, transcript texts and analysis results — 30 days, then deleted automatically; expired files are cleaned up every 30 minutes. Usage logs — for as long as they are needed for billing and reporting. The account — for as long as you use it.

If you dispute an assessment, the complaint, your score, original assessment, decision and a copy of the call transcript are retained until you or the Service owner delete them, including beyond 30 days. Labeling examples from accepted or partially accepted disputes have the same retention and are used for your account only. Up to 1000 recent disputes are retained per account, with the transcript copy limited to its first 2000 characters; exceeding the limit deletes the oldest dispute. You can delete a dispute and its example in the client area or through DELETE /v1/calls/{call_id}/dispute; DELETE /v1/calls/{call_id} also deletes the associated dispute and example.

For a record with analysis, DELETE /v1/calls/{call_id} with your API key deletes the stored audio, transcript, analysis result and job early. Usage-log metadata remains for billing. Deleting a record before the next delivery attempt stops its webhook; a delivery already in progress is not cancelled. Previously sent notifications and recipients' copies are not deleted by this request. You can have the account and its data deleted by writing to support@voicesscribe.com.

7. Changes and contact

The current version is published on this page, with the date shown above. Data questions — support@voicesscribe.com.