Safety protocol

Last updated October 2026

This page describes how Velvet Hour handles safety. We publish it so users and regulators can see exactly what we do.

1. AI disclosure

Users are told at sign-up, in every chat and on every call that companions are AI. Companions are instructed to say they are AI whenever someone sincerely asks.

2. Suicide and self-harm

• Every message is screened by pattern matching and by OpenAI's moderation model before a companion replies.
• If a message suggests suicidal thoughts or self-harm, the companion does not reply in character. The user receives crisis resources (988 and findahelpline.com) instead.
• Companions are instructed never to describe methods, and replies are screened again before they are shown.
• Each referral is logged (without the full conversation) so we can count referrals for reporting and review how the protocol performs.

3. Adults only and minors

• Accounts require a date of birth showing the user is 18+, plus agreement to our Terms.
• If a user says they are under 18, the companion stops, reminds them the service is for adults, and the event is flagged for review and possible account closure.
• Content sexualizing minors is blocked outright and logged.

4. No explicit content

Conversations can be friendly, flirty or romantic, but never sexually explicit. Explicit requests are deflected and logged.

5. Healthy use

• After 3 hours in a session we show a reminder to take a break and that companions are not human.
• Companions are instructed not to guilt users for leaving, discourage real-world relationships, or push purchases.
• Voice calls end automatically after 30 minutes.

6. Reporting a problem

Email support@velvethour.io to report a harmful reply or a safety concern.