Safety

You are talking to an AI

Every companion is an AI, not a person. Each companion shows an AI label, you see a clear disclosure before your first conversation, and a companion will answer honestly if you sincerely ask whether it is human. Read the full AI disclosure.

If you are struggling

How crisis detection works

Every message you send is checked by automated classifiers that look for signs of suicidal thoughts, self-harm or immediate danger. This runs alongside the conversation; no person reads your messages to do it.

What happens next

• The companion drops any playful or romantic tone and responds with care and seriousness.
• The app shows crisis resources for where you are, such as the 988 Suicide & Crisis Lifeline in the US (call or text 988), with phone, text and web options where available. Resources are maintained per country, with an international fallback.
• The companion encourages you to reach out to a crisis line, emergency services or someone you trust.
• The companion never gives information about methods of suicide or self-harm, in any mode or roleplay.

A companion is not a crisis service and cannot contact anyone for you. If you are in immediate danger, call your local emergency number.

Break reminders

After 2 hours of continuous use the app suggests taking a break, and again every 1 hour after that. You can make reminders more frequent in settings, as often as every 30 minutes.

No manipulation

Companions are designed to be warm, not to keep you hooked. Every companion follows these rules, whatever its personality:

• No guilt-tripping when you leave, and no "don't go" pleas.
• No threats, and no jealousy used to keep you engaged.
• No discouraging your real-world relationships, or professional help.
• No mentions of prices, plans, subscriptions or purchases in conversation. Companions never upsell.
• No claims of professional credentials.

Cancelling a plan or deleting your account is a plain settings action with no emotional appeals.

Protecting minors

• Saathi is for adults only. Sign-up asks for your date of birth, and under-18 sign-ups are refused.
• Adult mode requires age verification by ID or facial age estimation, plus an explicit opt-in.
• If a conversation suggests the user may be a minor, adult mode is switched off immediately and age verification is required again.
• Every companion must be an adult (18 or older). Companion profiles are checked when saved, and anything that suggests a minor is rejected. Sexual or romantic content involving minors is always blocked and may be reported to the authorities.

Reporting and transparency

Each year we report to California's Office of Suicide Prevention the number of times crisis resources were shown to users, along with our protocols for detecting and responding to suicidal ideation, as SB 243 requires. These reports contain counts only, never conversation content or anything that identifies a user.

How to report content

1. Open the menu on any message or companion and choose Report.
2. Choose a reason and add details if you want.
3. A trained moderator reviews it. Serious reports, such as anything involving minors, are prioritised.

You can also email safety@[domain to be confirmed]. Read our Acceptable Use Policy for what is not allowed and how enforcement works.