Safety & Crisis Protocol

Last updated July 8, 2026 · This page describes how Impela responds when a conversation shows signs of crisis.

What Impela is — and is not

Impela’s coach is an artificial-intelligence computer program, not a human being. It is a coaching tool — it is not therapy, not medical or mental-health care, and not a crisis service. Every account acknowledges this at signup, and the product reminds you that you are talking with AI at the start of coaching and at regular intervals during long conversations.

How we detect signs of crisis

Every message in a coaching conversation passes through two separate safety checks, automatically and in real time:

• A fixed rule, enforced by code — not left to the AI’s judgment — that catches clear statements of self-harm or suicidal intent, with care taken to tell real signals apart from figures of speech (“this project is killing me”).
• A second check that looks for warning signs even when nothing is said outright — hopelessness that builds message to message, a sense that there’s no future, giving things away, a sudden eerie calm — including signs that add up across separate conversations.

This screening is automated and runs inside your live session — not by a person. No Impela employee reads your conversations, and your data is encrypted in transit and at rest.

What happens when a signal fires

• The coach stops normal coaching. It will not problem-solve or continue the topic that raised the signal, and it will never share information that could help someone self-harm.
• A crisis-resources notice appears in the conversation with the contacts listed above, and the coach encourages reaching a human — a crisis line, a professional, or a trusted person in your life. If a signal was ambiguous, the coach asks you directly and plainly rather than guessing.
• We count each crisis-referral notice we show (a number only — never the content of your conversation) so our safety practices can be reported and audited as required by law.
• For 72 hours afterward, all marketing and engagement messages (reminders and summaries) are automatically silenced.
• If concerns were raised but not resolved, the coach checks in again the next time you talk, rather than letting it drop.

Teens (ages 13–17)

A teen can use Impela only when a parent or guardian invites them and the teen agrees. There is no way for a child to sign up alone. Under 13, no one can join — if we learn a user is under 13, we remove their data.

How the teen program keeps young people safe:

• A parent invites, the teen accepts. Both have to say yes before any teen coaching begins.
• A teen’s conversations are private. Parents cannot read their teen’s coaching, and neither can we. A parent sees only what the teen chooses to share.
• Parents are alerted to serious safety risk — never the words. If our safety checks detect signs of real danger, we alert the teen’s parent (and any trusted adults the parent has added) right away. The alert says to check on the teen and includes crisis resources. It never includes what the teen said.
• The coach reminds teens it’s AI. On a regular schedule during longer conversations, the coach reminds a teen that it is a computer program, not a person, and to take a break.
• Some limits are enforced by code, not left to AI judgment. Certain topics that aren’t appropriate for a teen are turned away by a fixed rule, with a plain explanation — not by the AI’s discretion.
• Crisis help comes first, and nothing is ever sold in a hard moment. If a teen is struggling, the coach points to real human help (a trusted adult, 988, and the Crisis Text Line) before anything else. A teen never sees a price, an ad, or a checkout — ever.

Questions

Contact support@impela.ai. If you believe this protocol failed you or someone you know, please tell us — reports are reviewed with priority.