Safety

Last updated: October 5, 2026

Helplines we show

• United States: 988 Suicide & Crisis Lifeline — call 988, text 988
• Canada: 9-8-8 Suicide Crisis Helpline — call 988, text 988
• Australia: Lifeline — call 13 11 14, text 0477 13 11 14
• United Kingdom, Ireland: Samaritans — call 116 123
• New Zealand: 1737, Need to talk? — call 1737, text 1737
• France: 3114 — call 3114
• Germany: TelefonSeelsorge — call 0800 111 0 111
• Spain: Línea 024 — call 024
• Brazil: CVV — call 188
• Israel: ERAN — call 1201
• Netherlands: 113 Zelfmoordpreventie — call 113
• Everywhere else: Find A Helpline

Each number was checked against the service’s own website. The country is worked out from your connection, so if you are travelling or using a VPN, the line shown may not be the one where you are.

Our crisis protocol

What we look for

Every message you send to a character — in the playroom, in group chats and in the previews on our homepage — is checked for signs that you might be thinking about suicide or self-harm. So is text you write into your Memory Journal and the descriptions of characters you create. The check has two parts:

• An automatic check for explicit statements — wanting to die, intending or planning to end your life, having the means, hurting yourself, or saying goodbye — written for English, Spanish, Portuguese, French, German, Hebrew, Arabic and Russian. It runs on every message, works without any outside service, and cannot be switched off by a technical fault.
• An AI screener, in any language, that reads the message the way the Columbia Suicide Severity Rating Scale (C-SSRS) screening questions do: a wish to be dead, thoughts of suicide, thinking about a method, intent, a plan, and recent or ongoing actions, along with self-harm. It also tells apart a story you are writing, a song lyric or a figure of speech (“this traffic is killing me”) from something about you.

If the screener is unavailable, the automatic check still applies on its own, and when it is unsure it leans towards treating a message as a sign that you may be struggling.

What happens

• You are never cut off. The character always answers, and free-message limits are lifted for the rest of that conversation’s difficult moment.
• A card appears in the conversation with a crisis line for your country, with buttons to call or text it, and a link to this page. It stays in the conversation.
• The character steps out of any roleplay or story. They speak plainly and warmly, take what you said seriously, ask whether you are safe right now, remind you that they are an AI, and encourage you to contact the line on the card. They never give information about methods.
• For the next 6 messages the character stays gentle: no flirting, no romance, no going back into a scene.
• If a message sounds like you are struggling but shows no sign of risk, the character checks in gently and a small “Need to talk to someone?” link appears, pointing here.

What we keep, and what we don’t

• A message flagged as a crisis is never added to a character’s memory. It does not appear in memory callbacks or in weekly recap emails, and it is left out of conversation summaries. The same goes for the character’s replies while the conversation is in crisis mode.
• For each crisis card or support link shown, we record an anonymous count: when it happened, which kind it was, the country and which part of the service. No message text, no account and nothing else that could identify you is stored with it. The law in California requires us to report these counts once a year to the state’s Office of Suicide Prevention.
• Nothing about a crisis is sent to our analytics or advertising tools.

What this is not

Svila.io is not a crisis service, a medical service or a substitute for professional help. Our checks are automatic: they can miss things and they can misread them, nobody watches conversations in real time, and we cannot contact emergency services for you. If you or someone else is in danger, call your local emergency number.

You are talking to an AI

Every character on Svila.io is an AI, not a person. Every conversation shows an “AI character” label next to the character’s name, and a notice in the conversation at the start of each session and again after every three hours of continuous chatting. Characters do not give medical, legal or financial advice, predictions about pregnancy, illness or how long anyone will live — they will point you to a qualified professional instead.

18 and over only

Svila.io is for adults aged 18 and over. AI companions may not be suitable for some people, including minors. If you notice that conversations here are making you feel worse, more isolated, or harder to step away from, it is a good moment to take a break and talk to someone you trust.

Report a problem

If a character said something harmful, or the protocol did not work the way this page describes, please tell us through our contact page. See also our Terms and Community Guidelines.