AI Voice Safety and Acceptable Use Policy
This policy sets the rules for every way SaydiVoice can produce a voice — text to speech, voice cloning, voice design, dubbing, voice changing, the API and the MCP connector. It is part of our Terms of Service and applies to everyone who uses the service, including through a reseller, an integration or a product built on our API.
1. Scope and definitions
Content means anything you submit (text, audio, reference recordings, transcripts) and anything the service produces for you (generated audio, dubbed tracks, cloned voices). Voice subject means the natural person whose voice a recording captures, or whose voice a cloned voice imitates. Cloned voice means any voice profile created from a reference recording, whether registered with us or supplied with each request.
This policy covers the conduct, not only the words. Attempting to do something this policy forbids is itself a breach, whether or not the attempt succeeds — for example trying to defeat a consent check, a usage limit or a safety control.
2. What you may do
You may create and publish synthetic voice where you have the right to use the voice and where listeners are not deceived about something that matters. Typical legitimate uses:
- Your own voice, for narration, courses, podcasts, accessibility or personal projects.
- A voice you are licensed to use — a voice actor under contract, an employee or presenter with written authorisation, a creator who has agreed to your terms, a licensed catalogue voice.
- Organisational voice where your organisation has granted you the right to produce material in its name.
- Clearly fictional voices that are not designed to be recognised as a specific real person, and do not exploit a real person's reputation.
- Dubbing, localisation and accessibility of content you have the right to reproduce, subject to the authorisation of the original performers where their voices are being imitated rather than translated.
In every case the authorisation has to cover the use you actually make of it: a consent for a internal training video is not a consent for a paid advertisement, and a licence that ends next month does not authorise a campaign that runs for a year.
3. What you must not do
The following are prohibited on every surface of the service.
3.1. Voice and identity
- Unauthorised voice cloning — creating a cloned voice from a recording of a person who has not given you explicit, informed permission.
- Impersonation — using a cloned voice to pass yourself off as another person or organisation, or to imply their endorsement, sponsorship, affiliation or approval.
- Fabricated statements — putting words into a real person's mouth that they did not say, in a way a listener could reasonably believe.
- Unauthorised celebrity or public-figure replication — reproducing the voice of a well-known person without authorisation, including for commentary, parody or satire where listeners could take it as genuine.
- Identity theft and the use of synthetic voice to obtain credentials, bypass verification or access an account.
- Impersonating a deceased person in a way likely to cause distress or harm to their family, or that misrepresents what they said in life.
3.2. Fraud, deception and influence
- Fraud and financial scams — including voice-based requests for money or credentials, fake emergency calls, and impersonation of a bank, a public authority or a colleague.
- Deepfake deception — presenting synthetic audio as a genuine recording.
- Misleading political communication — synthetic voice presented as a real candidate or official, or used to suppress participation in an election or public consultation.
- Misleading financial or health communication — synthetic voice attributed to a real adviser, doctor, regulator or company, or giving investment, medical or legal advice that listeners could act on as if a qualified person had said it.
- Fabricated evidence — synthetic audio presented as a recording of a statement, confession, testimony or event for use in a legal, disciplinary, insurance or regulatory process.
- Extortion, harassment and intimidation, including synthetic audio used to threaten, coerce or humiliate.
3.3. Privacy, intellectual property and other people's rights
- Privacy infringement — cloning a voice from recordings obtained without the speaker's knowledge, or generating audio that exposes private information.
- Intellectual property infringement — using a recording, performance or voice you do not have the rights to use, or a voice that is protected as part of a character or brand.
- Voice profiling — using voice data to identify, categorise or profile a person, or to infer sensitive characteristics about them.
- Children — submitting a recording of a person under 18 to create a cloned voice, or producing audio designed to be recognised as a specific real child. Accounts are for people aged 18 or over.
- Circumventing our controls — evading consent checks, usage limits, rate limits, review queues or account restrictions, including by opening additional accounts or API keys to do so.
4. Transparency about AI-generated audio
You are responsible for complying with the transparency rules that apply where your content is published, and with the labelling duties of Vietnamese law for content that simulates a real person's voice. In practice:
- Make clear that a voice is synthetic wherever a listener could otherwise believe it is the real person — in the description, the caption, the surrounding context or an audible notice, whichever fits the medium.
- Keep any machine-readable provenance or marking that the service attaches to audio. Do not remove, alter or conceal it in order to disguise that the audio was generated.
- Do not describe synthetic audio as a recording of a real event.
- Where the law of the place you publish in requires a specific form of disclosure, that form governs.
These duties sit alongside, and do not replace, the authorisation the voice subject gave you.
5. Enforcement
We investigate reports and risk signals, and we can act without notice where the risk is serious. Depending on what we find, we may:
- Disable one cloned voice, so that it stops rendering audio on every surface while it is reviewed.
- Refuse or stop individual generations, or require you to evidence your authorisation before further use.
- Suspend voice cloning on an account while a review is open, leaving the rest of the account usable.
- Suspend or terminate the account for a serious or repeated breach.
- Preserve and disclose information where the law requires it, or where it is necessary to prevent or respond to a serious risk of harm.
We look at context before acting: the purpose of the content, whether the voice subject can be identified, whether listeners could be deceived, and whether harm is already occurring. A satirical voice paid for by the subject is not the same thing as an unsolicited clone of a public figure used to sell a financial product, even though both are synthetic.
Consequences that follow from a decision are applied in the service, not only in a message: a disabled voice cannot be used through the studio, the API or the MCP connector, and a suspended account cannot register new voices or render from cloned ones.
6. Reporting a problem
If your voice has been cloned or used without your permission, or you have found content on SaydiVoice that breaches this policy, use our abuse report form. It is open to anyone — you do not need an account. You can also write to contact@saydi.ai.
Each report gets a ticket reference. What happens next, and the timescales we commit to, are set out in the voice consent policy and on our Trust & Safety page.
Deliberately false or abusive reports are themselves a breach of this policy.
7. Appeals
If we disable a voice or restrict your account, you can ask us to look again. Reply to the notice we sent, or write to contact@saydi.ai, quoting the ticket reference and explaining why you believe the decision was wrong — usually with the authorisation you hold. A different person from the one who made the first decision reviews the appeal where staffing allows, and we tell you the outcome with reasons.
An appeal does not suspend the enforcement while it is being considered, because the risk that justified the action is still present until the review says otherwise.
8. Changes to this policy
We may update this policy as the law, the risks or the service change. The date at the top of this page is the date the current version was published; a material change is announced in the studio or by email before it takes effect, where we can reasonably do so. Continuing to use the service after a change means you accept the updated policy.
Legal sources cited
The instruments below are the legal bases cited in this document. The links point to the Government legal portal and the signed original.
- Law on Artificial Intelligence No. 134/2025/QH15 — passed by the National Assembly on 10 December 2025, in force from 1 March 2026
Government legal portal · Signed original (PDF) - Law on Personal Data Protection No. 91/2025/QH15 — passed by the National Assembly on 26 June 2025, in force from 1 January 2026
Government legal portal - Decree 356/2025/ND-CP detailing the Law on Personal Data Protection — issued by the Government on 31 December 2025, in force from 1 January 2026 and replacing Decree 13/2023/ND-CP
Government legal portal
See the provision mapping and the full text of every cited provision