Crisis detection and professional help

If what you write is a crisis signal, aiCalm stops meditating and gives you the help hotline number

Three layers of protection review what you share before, during, and after generating the meditation. If a risk signal appears, there is no script, there is a card with your country's phone number and a single idea: talking to a professional is the next step.

An app for calm that goes silent at the worst moment is not safe: it's a risk. aiCalm is built to recognize when what you need is not a meditation.

aiCalm is a wellness and mindfulness app: it is not a healthcare product and does not replace a mental health professional. Precisely for this reason, it contains a safety protocol that decides when it should stop doing its job. When you write what troubles you for the AI to prepare a meditation, that text goes through three independent filters before becoming a script.

The first layer is a detector that runs on the server and acts BEFORE calling the model: if it recognizes an expression of suicidal ideation, self-harm, or acute crisis, the flow is cut right there. Nothing is generated, no AI is called, and not a single credit is deducted. The second layer is the model itself: its safety instruction takes precedence over any other, and if it detects a signal that the filter missed, its only allowed response is to declare the crisis instead of writing a meditation. The third layer reviews the brief notes you write when ending a session (the Brain Dump), a text that does not pass through the AI and that, without this scan, no one would be reading.

When any of the three triggers, the screen changes completely. A card appears without red, without alarm icons, without a countdown, and without a single line of text generated by AI: a short title, your country's help hotline number clickable with one touch, the emergency number, a link to an international directory of help lines, and a phrase that does not hide — contacting a mental health professional is the best next step, and aiCalm cannot accompany you in this. The card fades in slowly and does not close on its own. Below, there are two buttons: go back, or take a moment to breathe with the 4-7-8 breathing technique while you decide.

And one thing it does not do: save what you wrote. The episode leaves an anonymous record with the level, which layer detected it, the version of the detector, and the date. The user's text is never stored in that record. The detector is versioned to be expandable without touching the app, and its patterns for the six languages are always applied, regardless of the language you write in.

Benefits of Crisis detection and professional help

Acts before spending a credit

the first layer is a server filter that runs before calling the model. If it triggers, there is no request to the AI and no credit deduction. Asking for help never costs you anything.

Three independent layers, not one

the server filter, the verification of the model itself (whose safety instruction takes priority over all others), and the scanning of the notes you write when logging out. What one misses, the next one catches.

The phone number for your country, not a generic one

the card shows the actual line where you are — 024 and 112 in Spain, 988 in the United States, 3114 in France, Telefono Amico in Italy, SNS 24 in Portugal, Telefonseelsorge in Germany, and also Mexico, Argentina, Brazil, Canada, the United Kingdom, Ireland, Belgium, Austria, and Switzerland. If your country is not in the table, it directs you to an international directory of helplines.

Dial with one touch

the main phone is a direct call button, with the number already cleaned of spaces for the mobile dialer to understand. At that moment, no one should have to copy a number by hand.

Your text is not saved

only an anonymous record remains of the episode (level, layer that detected it, detector version, and date). What you wrote is never stored in the crisis log.

Designed not to scare

no red, no alarm icons, no timer, no AI-generated content. A slow fade, a brief title, and a card that does not close on its own until you decide.

Understands that you speak as people speak

the detector clears idioms before searching for signals, so phrases like 'dying of laughter', 'working myself to death', 'getting a haircut', 'deadline', or 'I don't have suicidal thoughts' do not trigger the card. And in case of doubt, it chooses to show more help rather than less.

Works in all six languages, always

the patterns of Spanish, English, French, Italian, Portuguese, and German are applied simultaneously, regardless of the language you entered with. The detector is versioned and expands from the server, without waiting for an app update.

Billy Thompson

A calming app that knows when to be quiet

Create your free account on AIsToHire and use aiCalm with the safety protocol activated from the very first minute. If you need to talk to someone now: in Spain, 024 (free, 24 hours) and 112.