Effective 31 August 2026 · version 2026-08-31.1
Voice and Third-Party Content Notice
This notice explains what happens when you send SAYDO a voice note, audio recording, message, image or document, including material created by or containing information about another person. It forms part of the Service Terms and should be read with the Privacy Notice.
Before you send content
- Send content only if you own it or have authority and a lawful basis to use it.
- Tell speakers and other affected people about the processing, and obtain permission where the law or your relationship requires it.
- Do not use SAYDO for covert recording, surveillance or monitoring.
- Transcription and AI can make mistakes. Review names, figures, dates, instructions and other important facts.
- SAYDO processes only content delivered to its official WhatsApp number; it does not scan your device or scrape unrelated personal chats.
1. What happens to a voice note
SAYDO receives media through the official Meta WhatsApp Cloud API after you choose to send or forward it to the official SAYDO number. The service checks the sender, message identity, media type, size and file structure before supported audio is downloaded into private temporary processing storage. Unsupported, malformed, mismatched or oversized media is rejected without being sent to the transcription provider.
Supported audio is sent to the configured transcription service. SAYDO then applies a bounded polishing step intended to return a concise, readable version in the speaker's original language and first-person voice. That step may remove filler, stutters, false starts and repetition and may safely reorder wording, but is designed to preserve material names, numbers, dates, times, amounts, requests, decisions, qualifications, commitments and uncertainty. If SAYDO's fidelity checks cannot produce a reliable result, the job may be rejected instead of returning a guessed transcript.
2. Direct and forwarded audio behave differently
SAYDO uses the forwarding information delivered by WhatsApp for the incoming message:
- Direct audio: after transcription, the transcribed words are treated as your AI request. If the audio is a WhatsApp reply to an earlier eligible message, SAYDO may include that sender-owned quoted context when answering.
- Forwarded audio: SAYDO returns the polished transcript itself. It is not automatically treated as an instruction to the AI. Available follow-up options can then be used to ask AI, summarise, extract tasks or prepare a draft.
This routing is based on provider metadata, not on SAYDO listening for a phrase such as “please transcribe”. Older queued messages created before routing information was stored may use the conservative transcript-only path. A transcript returned to WhatsApp is sent as plain text without decorative opening and closing quotation marks.
3. Accuracy and appropriate use
Speech recognition and generative AI are probabilistic. Background noise, accents, overlapping speakers, names, technical terms, low volume, compression and poor recordings can produce missing, altered or uncertain text. Polishing can improve readability but cannot make unclear audio certain. An output can therefore be fluent and still be wrong.
You must check important output against the original content before using, sharing or acting on it. Do not use SAYDO output as the sole basis for a legal, medical, financial, employment, credit, safety or other decision that significantly affects a person. SAYDO is not an emergency service, a certified transcription service or a substitute for a qualified professional.
4. Your authority and the rights of other people
By submitting content, you confirm that you have the rights, authority and lawful basis needed to let Amax Digital (Pty) Ltd and the service providers identified in the Privacy Notice receive, transmit, transcribe, analyse, store for the disclosed period and return that content. This includes copyright or confidentiality permission and any privacy, workplace, interception or recording consent required for other speakers and people identified in the material.
Recording and communications laws differ by country and context. Permission to participate in a conversation does not necessarily permit you to record, forward, analyse, disclose or retain it for every purpose. If you are acting for an organisation, you must also follow its workplace, customer, confidentiality, information-security and records policies. SAYDO's technical ability to process a file is not legal approval to submit it.
Do not submit covert recordings, unlawfully intercepted communications, stolen documents, confidential material you are not authorised to disclose, or content used to stalk, profile, harass, discriminate against or secretly monitor another person. Do not submit sexual content involving a child or any content that exploits or endangers a child.
5. Sensitive information and secrets
A recording or document may reveal identity, health, finances, beliefs, employment, location, relationships or other sensitive facts. Submit only what is necessary for the task. Do not send passwords, one-time codes, private keys, authentication tokens, full payment-card or bank-account details, government identity numbers or information whose use through WhatsApp is prohibited by law or provider policy. Redact unnecessary information before sending where practical.
6. Storage, providers and deletion
SAYDO removes its local temporary media directory after a processing attempt completes, including after a handled failure. The original WhatsApp message remains subject to your and Meta/WhatsApp's own retention controls. The transcription/AI provider may process content and keep provider-side records under the data-control terms described in the Privacy Notice. SAYDO may retain the resulting transcript, AI response, encrypted conversational record, quoted reference and operational metadata for the disclosed periods when memory or history is enabled.
SAYDO does not use submitted voice to identify you biometrically and does not build a voiceprint as a product feature. It does not continuously listen through your microphone, access unrelated WhatsApp chats, monitor personal conversations or import your chat history. Only messages intentionally delivered to the official SAYDO service enter this workflow.
Available controls include MEMORY OFF, RESET, CLEAR MEMORY, DELETE HISTORY and verified account deletion in the dashboard. Each control has the scope explained when it is used. Removing a number or sending STOP prevents future WhatsApp processing and cancels eligible queued work, but does not by itself delete completed history. Provider, payment, audit, security, dispute, backup or legally required records may remain for the periods in the Retention Policy.
7. WhatsApp consent and stopping service
Your WhatsApp service consent covers account, verification, usage, credit, billing and user-requested processing messages on the number you link. You can send STOP to withdraw that consent and suspend future processing for the number. Reconnection requires fresh dashboard consent and a new number-control proof. Withdrawal does not reverse processing already completed before it was received.
8. Reports, questions and rights
If you believe content was submitted without authority, or want to request access, correction, deletion, restriction or objection, contact info@saydo.co.za. Include enough information for SAYDO to identify and verify the request, but do not email passwords, authentication codes or unnecessary identity documents. Privacy rights and complaint routes are explained in the Privacy Notice and access-to-records requests in the PAIA Manual.
This notice describes the current direct-versus-forwarded audio workflow. A future feature that changes the purpose of voice processing will require corresponding product, consent and notice review before release.