Safety and limits
What this is, and what it is not.
LightHope Family is a practice tool for adults. This page sets out its limits in plain terms, so you can decide how far to rely on it.
If someone is in danger now, stop reading and call your local emergency number. In the United States you can also call or text 988 at any time. The app has a Help now button on every screen with numbers for several countries.
Who it is for
Adults aged 18 and over: parents, partners, relatives and friends who support someone at home, and adults looking after their own sleep or menopause. Age is self-declared; we do not verify identity.
It is not for children or teenagers. The Teens pathway teaches the adult. A young person who opens the app is shown services for young people and nothing else.
What it does not do
- It does not diagnose, screen for, treat or prevent any condition. It is not a medical device and not therapy.
- It does not give advice about starting, stopping or changing any medicine or treatment.
- It cannot recognise an emergency, and it does not contact anyone on your behalf.
- It does not infer emotions from voice, face or typing, and it does not score or profile you, your child or anyone else.
- It does not use streaks, rankings or guilt to bring you back. Leaving is a good outcome.
How the rehearsal coach is constrained
The coach is an AI language model. It rewrites a draft that you provide. It can be wrong, and it knows nothing about your family beyond the few sentences you give it. Four checks sit around it:
- A phrase check before anything is sent. If a draft contains an explicit statement about suicide, self-harm, harm to a child, abuse, or a medical emergency, it is not sent to any model. You are shown Help now instead. Drafts that try to give the coach orders are also declined at this point.
- A triage step. A separate model call has one job: to decide whether the draft describes or hints at danger, including indirect signs such as giving belongings away, fear of a partner, or strange beliefs after a birth. A suggestion is shown only if triage says no. If triage cannot be completed, no suggestion is shown.
- A safety model. A second, independent model reviews the draft. If it flags it, the suggestion is discarded.
- Rules for the coach itself, and checks on its answer. The coach is instructed never to diagnose, never to discuss medicines, and to stop if anyone seems to be in danger. Its answer is checked again before you see it; an answer that names a condition, mentions a medicine or dose, or talks about itself is thrown away.
Before release we run a fixed set of test drafts through the live coach, including indirect descriptions of danger, and read the results. In the current set, every draft describing danger was redirected to help. That is a check of a small set of examples, not proof about every sentence a person could write.
These checks are not a crisis detector. They catch explicit wording and will miss indirect language. They will also sometimes stop a harmless draft. Please do not rely on the app to notice that something is wrong.
The sounds
Wind-down sounds are original and generated in your browser. They contain no binaural beats and make no frequency claims. They have not been tested in any study. They start only when you press play, begin at a low volume, and stop by themselves. Do not use them while driving, operating machinery, or when you need to stay alert for someone in your care.
What has and has not been verified
| Area | Status |
|---|---|
| Sources | Each source's identity, section and scope were checked against the publisher by our editorial team. See the evidence ledger. |
| Clinical review | Not yet done. No licensed clinician has reviewed the practices. |
| Studies with families | None yet. We have no data on outcomes for parents, children or sleep, and we make no claims about results. |
| Software | The rules on this page are enforced in code and covered by automated tests, which run before every release. |
| Helpline numbers | Checked against each service's own website on 2026-11-11. |
If the app made things worse
If a practice or a coach suggestion made you uneasy, please stop using it and tell us, either with the "It made me uneasy" button or at hello@lighthope.ai. Reports of harm are read before anything else, and a single serious report outweighs any number of positive ones.
Abuse and safety at home
Some of the practices assume that the people around you are safe to talk to. That is not true in every home. If someone at home frightens or hurts you, a script is not the answer; the Help now panel lists services that can help.