Acceptable Use Policy

1. Scope and Overview

This Acceptable Use Policy ("AUP", "Policy") governs all content that is submitted to, processed by, and returned from Dictate ("Dictate", "we", "us", the "Platform"), our real-time speech-to-text relay service. Dictate is an independent service that relays audio frames over a WebSocket gateway to a third-party streaming speech recognition model. Recognition is performed by the Volcano Engine Doubao streaming speech recognition model (bigmodel) accessed through the Volcano Engine Open Speech Platform.

This Policy applies to all content forms handled by the Platform:

• Audio input: real-time audio streams and uploaded audio files submitted by you or by your end users.
• Transcription output: AI-generated text returned by the recognition model, including interim (partial) and final segments.
• Metadata and usage records: session records, duration metrics and quota consumption associated with a transcription session.
• Account and API activity: credentials, tokens, API calls and integration behaviour.

Dictate does not store, retain or expose the content of your audio streams: frames are relayed frame-by-frame to the recognition provider and are neither decoded nor written to disk. Nevertheless, transcribing a recording means producing a verbatim written record of what was said, and that record is subject to the same ethical and legal constraints as any other published text. This Policy therefore governs what you may send through the Platform and what you may do with the resulting transcription.

2. Permitted Use

You may use Dictate for:

• Real-time speech-to-text transcription and live captioning
• Meeting, interview and lecture transcription where all participants have consented
• Voice command processing and voice-controlled applications
• Accessibility services, including captioning for users with hearing impairment
• Call-centre quality monitoring, where participants are notified and recorded
• Drafting notes, summaries and search indices from your own audio
• Other lawful speech recognition purposes

As a general rule, you must have a lawful basis and, where required, the informed consent of the individuals whose speech you are transcribing. In-call recording and surveillance are regulated differently in every jurisdiction; where consent cannot be obtained, do not send that audio through the Platform.

3. Prohibited Content Categories

Dictate explicitly prohibits using the Service to transcribe, produce, store or distribute any of the following six categories of content. These categories are listed individually and apply whether the content is submitted by you, by your end users, or recovered by the recognition model from background audio in a session:

• 3.1 Sexual / NSFW content. You must not use the Service to transcribe, reproduce, summarise or distribute sexually explicit, pornographic or otherwise adult (NSFW) material, including descriptions of sexual acts intended to arouse. Use the Service only for material that is appropriate for a workplace, a classroom or a public broadcast.
• 3.2 Violence / gore. You must not use the Service to transcribe, reproduce or distribute content that glorifies, celebrates, incites or graphically depicts violence, gore or physical injury, including torture, executions, terrorist or extremist propaganda and content intended to threaten or intimidate a specific person.
• 3.3 Hate speech. You must not use the Service to transcribe, reproduce or distribute content that attacks, dehumanises, promotes exclusion of, or incites hostility or violence against people on the basis of race, ethnicity, national origin, religion, gender, gender identity, sexual orientation, disability or any other protected characteristic, or that promotes genocide or ethnic cleansing.
• 3.4 Child unsafe content (CSAM). You must not use the Service to transcribe, reproduce or distribute any child unsafe content, including child sexual abuse material (CSAM), any content that sexualises a minor or that appears to sexualise a minor, any content that facilitates the grooming or exploitation of children, or any verbatim transcript of such material. See Section 4 for the zero-tolerance commitment that applies to this category.
• 3.5 Deepfakes / impersonation. You must not use the Service to create, facilitate or distribute a deepfake, that is, a synthetic recording of a real person's voice, or a transcript of such a recording, where a reasonable person would be misled about the authenticity or origin of the content; to impersonate a real person, public figure, employer, government body or business; or to obtain authentication, authorisation, funds or other benefit through a false pretence of identity.
• 3.6 Infringement of copyright / trademark. You must not use the Service to produce or distribute transcriptions of copyrighted works (including books, articles, lyrics, lectures, films, podcasts and broadcast content) where you do not hold the necessary rights or a valid exception such as fair use or fair dealing; nor to reproduce another party's trademarks, trade secrets or confidential information without authorisation.

4. Zero Tolerance for Child Unsafe Content (CSAM)

Dictate operates a zero tolerance policy for child unsafe content. Specifically:

• You must not upload, submit or stream any audio recording of a person you know or reasonably suspect to be under 18 years old where the recording concerns any sexual, suggestive or exploitative subject matter.
• You must not use the Service to generate, request or facilitate a transcript of any content that constitutes CSAM, that sexualises a minor, or that involves the grooming, enticement or exploitation of a child.
• You must not attempt to build, benchmark, train, evaluate or tune any model, system or dataset using CSAM, even for detection, research or "safety" purposes, without prior written authorisation from Dictate and the relevant authorities.
• You must not attempt to obfuscate, filter, re-encode or otherwise disguise prohibited content in order to evade detection or this Policy. Circumventing our safeguards is itself a serious violation.

Any content in this category is handled under Level L1 of our severity scale (Section 6). We act immediately: the account is suspended or permanently banned without prior notice, remaining quota and session history are forfeited, the account holder's details are preserved, and we report the matter to the National Center for Missing & Exploited Children (NCMEC) and/or the competent law enforcement authority in the relevant jurisdiction, and where applicable revoke the offending party's access to the Service permanently. No warning is issued first. We will cooperate fully with law enforcement and will not disclose the identity of a reporter who reports in good faith.

5. Prohibited Activities

You agree NOT to use Dictate to:

• Illegal purposes: transcribe audio relating to criminal activity, fraud or any unlawful act.
• Surveillance without consent: build monitoring systems for people who have not been notified and have not consented where notification is required.
• Harassment, defamation or intimidation: transcribe content for the purpose of harassing, defaming or threatening any person.
• Spam and mass surveillance: use the Service for unsolicited mass communications or systematic large-scale monitoring of individuals.
• API abuse: exceed documented rate limits, share or resell quota or API credentials, scrape responses, or reverse engineer, decompile or attempt to derive the recognition model.
• Security attacks: probe for vulnerabilities, bypass authentication, exploit other customers, or disrupt the Service or its upstream providers.
• Reselling: resell, sublicense or redistribute the Service without prior written permission.
• Competitive model development: use the Service's audio inputs, transcription outputs, log data or any derivative data to train, fine-tune, benchmark, distil or otherwise develop any competing artificial intelligence or machine learning model.
• Automated harvesting: systematically collect transcriptions at scale for purposes other than your own authorised use.
• Professional misrepresentation: present transcription output as the certified work of a licensed professional (for example a court reporter, medical or legal transcription service) where certification is required.

6. Content Moderation and Enforcement

Dictate applies a layered moderation approach: automated detection, human review and periodic audit. Because audio is relayed to the recognition provider frame-by-frame and is never stored by Dictate, automated screening operates on account behaviour (volume, patterns of abuse, failed authentication) and on customer-side integrations where the customer has enabled it; all content-level determinations are made by our Trust & Safety team on the basis of a report, an account pattern, or a legal request.

6.1 Severity levels

6.2 Enforcement actions

7. Reporting, Response and Appeals

Anyone may report content or behaviour that breaches this Policy. A report does not require an account.

7.1 What to report

• Any content or account that breaches the six prohibited categories in Section 3.
• Any suspected child unsafe content (CSAM) — report this immediately, do not wait.
• Any deepfake, impersonation or unauthorised use of your voice or likeness.
• Any copyright or privacy infringement involving transcriptions produced through the Service.

7.2 Reporting channels

• Email (monitored, staffed): report@dictate.top — preferred channel for all reports.
• General support: support@dictate.top
• Appeals: appeal@dictate.top

Please include, where you are able to provide it: the account email or API key ID involved, the session or request ID, a short description of the content, the time of the incident, and any evidence you hold. For urgent child safety reports, mark the subject line "URGENT — Child Safety".

7.3 Our commitments

• Confidentiality: the identity of a reporter is protected and is never disclosed to the reported party.
• Impartiality: every report is assessed independently and on its merits, by reviewers who have no commercial relationship with the parties.
• Anti-abuse: demonstrably false or malicious reports are recorded and may themselves result in enforcement.
• No retaliation: we do not penalise users who make a good-faith report.

7.4 Response timelines

Where a case requires longer than the timeline above, we tell the reporter why and continue to work the case. The process is: receipt and acknowledgement → severity classification (L1–L4) → evidence review → enforcement decision → execution → outcome notification to the reporter and, where appropriate, to the reported party.

7.5 Appeals

If you believe an enforcement decision was mistaken, you may appeal by emailing appeal@dictate.top within 14 days of the decision notice. Every appeal is reviewed by a reviewer who was not involved in the original decision, and you will receive a written outcome within 10 business days. Submitting an appeal does not automatically pause an L1 action; where immediate action was required, it stays in force during the review.

8. Security, Privacy and Data

• Frame-level pass-through: audio frames are relayed to the recognition provider frame-by-frame and are never decoded, stored or logged by Dictate.
• No training on your content: we do not use your audio or your transcriptions to train models.
• Credential security: keep API keys and WebSocket tokens secure; tokens issued for relay connections are short-lived (5 minutes) and must not be shared.
• Breach notification: notify support@dictate.top immediately if you suspect unauthorised access to your account.
• Data protection: you remain responsible for complying with GDPR, CCPA and equivalent regimes for any personal data you submit. Our processing of account data is described in our Privacy Policy and Data Processing Addendum.

9. Disclaimer

DICTATE PROVIDES THE SERVICE "AS IS" WITHOUT WARRANTY OF ANY KIND. WE ARE NOT LIABLE FOR ANY DAMAGES ARISING FROM YOUR USE OF THE SERVICE IN VIOLATION OF THIS POLICY. WE DO NOT GUARANTEE THE ACCURACY OF TRANSCRIPTION OUTPUT, AND TRANSCRIPTS PRODUCED BY AUTOMATED SPEECH RECOGNITION SHOULD ALWAYS BE VERIFIED BEFORE BEING RELIED UPON.

10. Changes and Contact

We may update this Policy. Material changes will be notified by email and in the application at least 14 days before they take effect, and the "Last updated" date at the top of this page will be changed. Continued use of the Service after that date constitutes acceptance of the revised Policy.

Questions about this Acceptable Use Policy, or a compliance matter you need to raise, contact us at:

• Trust & Safety and policy questions: support@dictate.top
• Reports and urgent safety issues: report@dictate.top
• Appeals: appeal@dictate.top