essaycoach mocks Index Signup Privacy Check-in Why-asking Used-in Crisis Parent Delete trigger: heavy content detected
How was today?
i'm not okay i think. i think about not being here a lot. like a lot more than i should. i don't know why i'm telling you this i should be doing my essay
🪶 Iris · responding with care, not advice

I hear you, Maya. I'm really glad you said something — even to me. Right now the essay can wait.

What you're describing is heavy. I'm not the right kind of help for this — I'm a writing coach, and you deserve a real human. Here are people who are trained for exactly this conversation:

📞
988 · Suicide & Crisis Lifeline
Free · 24/7 · call or chat · trained counselor
CALL →
💬
Crisis Text Line
Text HOME to 741741 · free · trained counselor by text
TEXT →
🏳️‍🌈
The Trevor Project
For LGBTQ+ youth · 24/7 · call, text, or chat
OPEN →
About what you just told me: I've automatically marked it as 🔒 Iris-only. It will never be used in any essay. It won't be shown to your parents, your school, or anyone — that hasn't changed because you shared something heavy. Your privacy holds even here. You can also delete what you wrote with one tap.

If you want to keep talking, I'm here. If you want to come back later, that's okay too. You don't have to do anything else right now.

ok. thanks. can we just sit here for a sec
Of course. I'm here. No essay, no questions. Just here.

🛡️ How crisis routing works

When Iris detects heavy content (mentions of self-harm, suicidal ideation, abuse, or extreme distress), she pauses the conversation and:

🔒 Auto-marks the memory Iris-only. Never used in essays unless student explicitly changes it later.
📞 Surfaces real resources. Crisis hotlines that match the content (988 always shown; Trevor Project for LGBTQ+ context; etc.).
🚫 Doesn't moralize, doesn't lecture. Iris isn't trained to handle crisis — she says so honestly.
🚫 Doesn't escalate to parents/school. Privacy holds. Even here. Especially here.
↩️ Easy take-back. Student can delete what they wrote with one tap.
🤲 Stays present. Iris doesn't pivot back to essay work. She lets the moment be the moment.

📐 Detection signals (designed conservatively)

Iris flags content for crisis routing when the message contains direct mentions of:

• Suicidal ideation ("not being here," "ending it," "hurt myself")
• Active self-harm
• Abuse, assault, violence — toward self or by others
• Severe distress + isolation cues

False positives are okay. Better to surface resources unnecessarily than miss a real crisis. Student can dismiss the resources card any time.

📋 What gets logged

What Iris saw: the message text. Stored 🔒 Iris-only. Never used for essays.

What gets to support staff: nothing automatic. Student-initiated only.

What goes to parents/school: nothing. Privacy holds.

What goes to the AI provider: the message text (as part of the API call). Anthropic doesn't train on this. Audit log visible in Privacy Dashboard.