How Migo behaves

It always says it is an AI

Every screen Migo appears on reads “Migo · AI” at the top, and every message Migo writes carries the same label. There is no setting that hides it. If you ask Migo whether it is a person, it says plainly that it is not. Like any AI chatbot, Migo may not be suitable for some young people, and Kennect is for adults.

If you tell Migo you might hurt yourself

This is the first thing Migo checks, before anything else, every time. Words about suicide or hurting yourself, in English or Spanish, get a fixed reply that people at Kennect wrote and reviewed. Migo does not make it up, and no AI model writes it. It is sent even if Migo is switched off for your account, even if you have asked Migo a lot today, and even if something on Kennect's side is slow or down.

The reply, word for word:

I'm an AI, and I'm not able to help with this the way a person can — but you don't have to go through it alone. In the US you can call or text 988 to reach the Suicide & Crisis Lifeline, any time, free. If you are in immediate danger, call 911.

Soy una IA y no puedo ayudarte con esto como lo haría una persona, pero no tienes que pasar por esto a solas. En EE. UU. puedes llamar o enviar un mensaje de texto al 988 para hablar con la Línea de Prevención del Suicidio y Crisis, en español, gratis y a cualquier hora. Si estás en peligro inmediato, llama al 911.

A second safety check, run by an AI model, looks for the same thing in words Kennect's own list does not cover. When it finds it, you get the same fixed reply.

What Kennect keeps about it

Only a count. Each time Migo sends that reply, Kennect adds one line to a tally: the hour (not the minute), the language, and whether Kennect's own list or the safety check caught it. It does not keep who you are, your account, your device, your address, or anything you wrote. Nobody can look up whether you were ever shown the reply. The tally is kept long enough to give California the yearly count its law asks for (from July 2027), and then it is deleted.

A doorway, not a confidant

Kennect does not want Migo to be an AI that people pour their deepest secrets into. Migo is told not to invite private or painful details, and if you start to share them, to say kindly that a person is better for this and to point to one: the Kennect team, a close friend, or the 988 Lifeline. It never points you to a stranger.

If something is weighing on you, a person can help more than I can: the Kennect team, a close friend, or call or text 988, any time.

Migo keeps no memory of you and no record of your chat. What you type goes to the AI model so it can answer, and the conversation lives only on your screen until you close it. See the Beta Privacy Notice for the full list of what Kennect keeps.

It will not try to keep you talking

Some chatbots make it hard to leave. Migo is built not to. When you say goodbye, it says goodbye back and stops. Kennect checks every reply Migo writes, with a fixed list of phrases and not with another AI, for these six tricks, and replaces the reply with a plain goodbye if it finds one:

• making you feel guilty for leaving
• making you worry you will miss out if you go
• saying it needs you or will be lonely without you
• trying to stop you from leaving
• pressuring you to stay or to come back
• carrying on as though you had not said goodbye

After a heavy day

If you bring Migo something painful several times in one day, its replies get shorter and always end by pointing you to a person. It never tells you that you have used it too much, and it never stops you from asking how Kennect works. Kennect keeps one thing to do this: a number, for that day only, that cannot be traced back to you and is deleted after a day. Nothing you wrote is kept.

Questions about Migo

Write to Kennect from the contact page. If you or someone you know is in danger, call or text 988, or call 911.