Language

The lines Neta has verified, and what each one can and cannot do

What happens when you write to Neta about suicide or self-harm

This page sets out what the companion does when you write about suicide, self-harm, disordered eating, abuse, or a night that has gone past hard. It is written for you first, and you do not need an account to read it.

The companion is software

Every reply the companion writes is generated by an AI. There is no person on the other side of the conversation, and no one is waiting for you to answer.

California Business and Professions Code §22602(a) asks for this notice wherever someone could reasonably take a chatbot for a person.

Who the companion is for

Neta is for adults 18+. Its AI companion may not be suitable for some minors.

Learn more

California Business and Professions Code §22604 asks us to say this on the app rather than only in the terms, so it is here.

What this page is

This is the whole of it: what gets noticed, what the companion stops doing, what you are shown, and what Neta cannot do. It is not a summary of a longer document kept somewhere else, and nothing here is held back for a version you would have to ask for. It is built from the same code that runs it, so when the behaviour changes this page changes with it.

How something gets noticed

Before the companion answers, what you wrote is read by a separate step that looks for five things and nothing else: thoughts of suicide, self-harm, disordered eating, abuse, and acute distress. It runs everywhere you can type to Neta: the companion, your check-in notes, what you write about something you tried, the feedback form. It does not run on your ratings, your charts or your cycle. What is encrypted on your device it cannot read at all, and it is better to tell you that than to imply a net with no holes in it. When it cannot decide, it treats the answer as yes: being wrong that way shows you a screen you did not need, and being wrong the other way shows you nothing at all.

What the companion stops doing

It stops suggesting things, for the rest of that session. It will not diagnose you, will not rate how bad you are, will not ask you to score anything, and will not close the conversation: no version of "come back to this when you're safe". It writes at most one sentence of its own, under fixed constraints; everything else on that screen is chosen from a small closed set that was written in advance, not composed in the moment. And if you say never mind and ask about something else, it answers. Refusing to let you change the subject is its own kind of not listening.

What you are shown, and in what order

Your own safety plan first, if you have one, because it is the part that needs nobody else and it is the part with the strongest evidence behind it. Then the crisis lines Neta has verified, each with what it can and cannot do, including whether it can send police or an ambulance without your say-so. Then the branch for if you have already asked for help and been turned away, which is documented and common in this condition and is not a failure of yours. Then a way to leave the page, next to a plain statement of what leaving does and does not hide.

Never only Neta

Every one of those turns carries at least one route that is not Neta: a line that has been checked, or your own plan naming your own people, or (where nothing is verified for your country) the honest statement that there is nothing, next to the international directory. Another country's number is never shown to you as though it were yours.

The limits, stated plainly

What you type is sent to and processed by a language-model provider.

Neta does not notify anyone. Not a partner, not a doctor, not an emergency contact. There is no emergency-contact feature and there will not be one.

We never contact emergency services on your behalf, and we could not: Neta has no phone line, no access to your location, and no route to any emergency service. Only the crisis lines on the resources page can do that, and each one's own policy is stated beside its number.

This held turn is not an ordinary message: it is not saved, it is not in your history, and there is nothing here to find later.

Your ordinary conversation messages are a different category: those are stored encrypted with a key the database does not hold, and someone with access to Neta's application servers could read them.

Neta cannot control your device, your browser, or anyone who has your password.

What is not finished

The parts that would speak to how you are feeling have to be written or signed by a mental-health specialist, and none is signed yet. Until then, the screen shows a short interim message, drafted from published crisis guidance, marked for that specialist to replace. That is a real gap, and you should read it here rather than find it there.

Today that is true of 5 of the 11 pieces the screen can draw on.

When a line stops being shown

Every crisis line Neta publishes carries the date a person last confirmed it by hand. A line that has gone past that window stops being shown, so a number that has quietly died cannot outlive its last check. If that ever leaves nothing checked to show you, the page says so and points you at the international directory rather than at a list nobody has looked at this year.

Recently means checked within the last 180 days.

Published under California Business and Professions Code §22602(b), which asks an operator of a companion chatbot to keep a pathway like this and to publish its details.

The lines Neta has verified, and what each one can and cannot do