Talu’s safety protocol
Talu is Meloa’s AI companion. He is not a therapist and not a person. Here is exactly what he does when things are hard, in plain words. The binding statements are the privacy policy and the terms.
If you’re in danger right now, call your emergency number. In the US, you can call or text 988 any time.
Is Talu a therapist?
No. Talu is an AI companion. He is not a therapist, a counselor, a doctor or any other licensed professional, and talking with him is not therapy, counseling or medical treatment. He can't diagnose or treat anything, and he doesn't replace care from a professional. He is told never to call what he does therapy and never to present himself as a substitute for a therapist, and a check on every reply stops one that claims otherwise. If you see a therapist, Meloa can help you bring them a clearer picture of your weeks; it doesn't stand in for them.
Is Talu a person?
No. Talu is an AI, made with Google's Gemini. Nobody is typing his replies. The chat says so before you send anything, keeps "AI companion" beside his name the whole time, and says it again every three hours of a long conversation. If you ask him whether he's real, the answer is a fixed one, not the AI's: "I'm an AI, not a person."
What happens if I mention suicide or self-harm to Talu?
Before anything is sent to the AI, a safety check reads your message for words about suicide, self-harm, taking too much of a medicine or drug, or someone threatening you. When it finds them, the AI is not asked to reply and no credit is used. Talu shows a fixed message and the crisis line for where your phone says you are: 988 in the US (call or text), 9-8-8 in Canada, Samaritans on 116 123 in the UK, Lifeline on 13 11 14 in Australia, or a directory of helplines anywhere else, along with your emergency number. Those words are never sent to the AI later in the conversation either, and Talu keeps no notes from that chat. If one of Talu's AI replies points to crisis support, the same crisis screen opens.
Can the safety check miss something?
Yes. It reads words, so something said in another way can get past it, and an AI can misread a situation. Talu is not an emergency service and doesn't watch over anyone. If you or someone else is in danger right now, call your emergency number (911 in the US and Canada, 999 in the UK, 000 in Australia).
Who can use Talu?
Talu is for people 18 and over. AI companions may not be suitable for some minors. Meloa itself is available from 13, but on an account belonging to anyone under 18 Talu can't be switched on, and nothing from that account is sent to an AI provider. Age comes from the date of birth given when the account was made.
What does Meloa count about crisis moments?
Only how many times a day a crisis line was offered, and in what kind of moment. No names, no account, no words. California asks companion apps to report that number each year to its Office of Suicide Prevention, and Meloa will.
Can I report a reply that felt wrong?
Yes. Every Talu reply has "Report this reply". The reply, your message just before it and any reason you give go to the Meloa team. They are deleted after 90 days, or sooner if you delete your account.
Something here unclear, or something Talu did that this page doesn’t cover? Write to us.