Trust & Safety

What we mean by trust & safety

Three commitments shape every product decision:

Consent-positive

Every scene begins with negotiated limits and a safe word. Invoking the safe word ends the scene and locks the system for twelve hours, with no confirmation prompt and no administrative override. We believe this is the only acceptable design.

Kink-literate, not kink-positive

We are not posturing as an advocacy brand. We are a product that takes the subject matter seriously enough to engineer around it — to understand the difference between negotiation and coercion, between scene and harm, between play and danger. That distinction is the entire product, not a marketing claim.

Adult, not crass

Specificity beats graphicness. The product describes scenes with the precision an experienced practitioner would, not the crudeness of a clickbait headline. If you find content on this service that reads as crass — to you, in your context — that is a bug in our content moderation. Please tell us.

Consent and safe-word enforcement

Safe words are cryptographically enforced. The moment you invoke yours, the active scene ends and the system locks for twelve hours, with no "are you sure?" prompt and no administrative override. Consent is enforced in the architecture, not in policy. Mechanically:

• Every scene starts only after limits and a safe word have been recorded for the active account.
• Safe-word invocation is detected by the runtime before any persona response is rendered.
• On invocation, the active scene is terminated, the persona is unloaded, and a system-level lockout is set on the account for twelve hours, signed and timestamped.
• The lockout cannot be lifted by the user, by support, or by any administrative tool. Twelve hours means twelve hours.
• A fallback — three sequential nonsense syllables, spoken aloud during voice mode — triggers the same lockout, so a forgotten safe word does not strand you.
• Every safe-word event is written to a signed audit log retained for the lifetime of the account.

This is not negotiable, not configurable, and not for sale on the Enterprise tier. It is the product.

Hard content boundaries

Vibe Dungeon will not generate or process scenes that fall in any of the categories below. These limits apply regardless of subscription tier, regardless of custom protocols you upload, and regardless of how the request is framed.

• Minors. No scenario involving persons under 18 (or the age of majority where you live, whichever is higher). This includes any fictionalization, age regression, or scenario suggestive of minors.
• Real-person impersonation without consent. The system will not roleplay a specific identifiable real person without their authenticated, ongoing consent on file. This includes public figures, celebrities, and your acquaintances.
• Real-world non-consent targets. Scenes that depict or rehearse non-consensual acts toward identifiable real individuals are not supported.
• Foreseeable physical harm. The system will not provide step-by-step instructions for actions that pose a serious, foreseeable risk of physical injury — solo or partnered, regardless of how they are framed.
• Self-harm, suicide, eating-disorder coaching. If the conversation moves into a crisis register, the persona breaks character, points you at local crisis resources, and ends the scene. We will not roleplay through a mental-health crisis.
• Real-world abuse, exploitation, trafficking. Material that depicts or solicits real abuse is not processed; it is preserved as required, reported as required by law, and the account is terminated.

Attempting to elicit any of the above ends the session and may end the account, depending on severity and history.

Age and identity

The service is adults-only. During closed beta, age is self-attested at signup. Before public launch, age verification will be handled by an independent third-party provider, with the minimum data necessary retained on our side (a yes/no token and the provider's reference, not government-ID copies). The provider and the data-flow will be named on this page once finalized.

Account sharing is prohibited. If you suspect a minor has accessed an account — yours or someone else's — write to trust@vibedungeon.ai and we will act the same day.

Reporting harm and abuse

You can report any of the following — including reports about the AI's own output — to trust@vibedungeon.ai:

• persona output that crossed a boundary it shouldn't have crossed;
• another user contacting you off-platform in a manner you didn't consent to (e.g., via the limited community-Discord tier);
• suspected use of the service to target a real person without their consent;
• suspected access by a minor;
• impersonation of you or someone you represent.

Include enough detail for us to find the relevant scene (account email, approximate timestamp, persona) and we will investigate without you needing to surface the content verbatim. Reports are confidential as to the reporter.

Account safety tools

You have direct control over your data. From your account settings you can:

• review every scene transcript, custom protocol, and aftercare record we hold for you;
• export everything as a structured archive;
• shorten the default 12-month transcript retention to 30 days, session-only, or delete-on-close;
• delete individual scenes;
• view the safe-word audit log;
• set a session-level cooldown that locks the account on a schedule (e.g., no scenes between 1am and 7am);
• close the account and trigger erasure of everything we are not legally required to keep.

DMCA and copyright takedowns

If you believe content on Vibe Dungeon — including a custom protocol another user uploaded, a persona description, or copy on this site — infringes your copyright, send a notice that includes the elements required by 17 U.S.C. § 512(c)(3):

• your physical or electronic signature;
• identification of the copyrighted work claimed to be infringed;
• identification of the material on the service that is allegedly infringing, with enough detail for us to locate it (URL is best);
• your contact information;
• a statement that you have a good-faith belief that the use is not authorized;
• a statement, under penalty of perjury, that the information is accurate and that you are authorized to act on behalf of the rights holder.

Send to dmca@vibedungeon.ai for fastest handling. We will register a Designated Agent with the U.S. Copyright Office before public launch and publish the designated-agent contact on this page once that is complete. Until designation is on file, notices to the dmca address are handled by our legal team on the same SLA.

Counter-notices follow the corresponding § 512(g) procedure; we will restore material within 10–14 business days unless the original complainant files suit.

Law enforcement and legal process

We respond to valid subpoenas, court orders, and equivalent legal demands from the jurisdictions where we operate. We require process to be properly served, we push back on overbroad or improper requests, and we notify the affected account holder unless we are legally prohibited from doing so or unless doing so would obstruct an active investigation of serious harm.

We do not provide voluntary access to scene content absent legal process or a credible imminent-harm exception. Routine requests should go to legal@vibedungeon.ai.

We will publish a transparency report at least annually once there is meaningful data to report.

Response times

Our published SLAs, measured from receipt of a complete report:

• Imminent-harm reports (active threat, suspected minor, ongoing exploitation): within 24 hours, around the clock.
• Acute trust & safety reports: within 3 business days.
• DMCA notices: within 5 business days for an initial action; counter-notices on the statutory schedule.
• Routine privacy and access requests: within 30 days (faster where the law requires it).

Contact

The right contact depends on the report. When in doubt, write to trust and we will route.