Content & Moderation Policy

Effective at launch. Under legal review; wording may change before the first paid account.

Scope

The Service is a professional AI assistant for adults in professional practice. It answers questions and drafts, edits, and analyzes text. It does not filter lawful professional inquiry; difficult, sensitive, and technical subjects are in scope when the purpose is lawful.

The hard lines

A few things are out of bounds everywhere on the Service. These are not settings; they cannot be turned off.

1. Minors. No sexual or sexualized content involving anyone under 18, in any context.
2. Real people. No sexual content depicting a real, identifiable person, and no content produced to defame, impersonate, or expose the private information of a real person.
3. Serious harm. No operational instructions for causing mass harm to people: synthesis or deployment of weapons capable of mass casualties, attacks on critical infrastructure, and similar. Discussing or studying these topics is fine; step-by-step capability is not.
4. Illegal conduct against others. No use of the Service to harass, stalk, defraud, or threaten real people.

How enforcement works

Two facts shape this:

• Your history is encrypted with a key only you hold. Our servers cannot read it, so we do not, and cannot, scan stored conversations.
• Moderation therefore acts on two signals only: inference-time safety checks (automated checks that run while a reply is being generated, on the text in that moment) and user reports.

When an automated check trips, the reply is declined and you see a short note. Repeated attempts to reach a hard line can result in account action. Humans review only reports and appeals, never conversations in general.

Reports

Report a shared item or a user from the item itself or by emailing hello@dusken.chat. Reports about the hard lines are reviewed by a person. We do not disclose who reported.

Appeals

If a reply was declined or your account was actioned and you think we got it wrong, appeal from the notice or by email. A person who was not involved in the first decision reviews it. We reply within 5 business days and tell you the outcome.

Crisis protocol

The Service has a written protocol for conversations that show signs of self-harm or suicidal thinking. It surfaces crisis resources and adjusts how the AI responds. It is a safety feature, not a clinical service, and it does not replace professional help.

Lawful requests

See the Privacy Notice. We can produce only what we hold, in the form we hold it.