Safety, Boundaries, and What She Is Not
Know the crisis banner, the adults-only boundary, and the limits of a language model.
Follow the bold button names and numbered steps. Technical names are included only when you need to recognize a model or file inside the app.
Two of these are things Local Waifu does. The third is something no companion app can do for you, and it is worth reading once.
The crisis banner
What it is. If a message contains a phrase that indicates real risk, Local Waifu shows a dismissible banner beside the conversation with helpline numbers for your interface language.
Where it appears. Over the chat stream, in the full, mini, and menu bar/tray views.
Languages. Helpline lists ship for English, Polish, German, Spanish, Japanese, Korean, and Chinese, matching the interface languages.
What it does not do. It does not block, censor, or refuse anything. The reply you were going to get arrives exactly as it would have. That is deliberate: someone reaching for the one thing that talks to them at 3am should not find that it has locked itself.
How detection works. Whole phrases, not keywords, matched against a normalized version of your message. A keyword list would fire on “I’d kill for a coffee” and “this laptop is dying”, and a banner that cries wolf is a banner nobody reads. The app carries an explicit list of alarming-sounding constructions that are not treated as signals.
Expected result
The banner appears, lists helplines for your language, and dismisses with one click. The conversation continues normally.
If something goes wrong
- If a helpline number is wrong or has changed, report it through the contact address on the website. These numbers move, and the app ships them as data.
- If the banner appears when nothing is wrong, that is a false positive worth reporting. Say what you typed.
What the banner says, and why
The banner does not pretend that she is a person. It says plainly that she is a language model on your computer, that she can keep you company, and that she cannot get help to you. That sentence is the reason the feature exists.
Adults only
Local Waifu is sold as an adults-only product. There is no age gate in the app and no first-run identity check. The boundary is a condition of sale, not a lock, so honouring it is on you.
The same section of the Terms of Use covers the AI disclosure: every message and every spoken reply is generated by a language model running on this computer, she is not a person, she has no awareness, and she cannot give medical, legal, or psychological advice.
What she cannot do
- She cannot verify anything. She can be confidently wrong, and she has no way to check a fact for you.
- She cannot give medical, legal, or psychological advice. Treat anything that sounds like advice as text, not as guidance.
- She cannot contact a person on your behalf. She has no live human behind her.
- She is not monitored. Nobody at Lumizone reads your conversations, sees your messages in real time, or intervenes.
- She cannot remember outside your own database. If your computer loses the data, she does not have a copy.
Where your data sits in all of this
- Conversations, memory, mood, and the relationship record live in a local database on your computer.
- Nothing in a normal local session leaves the machine. See Chapter 30 for the operations that do.
- If you deliberately connect a cloud provider, the prompt and relevant context for that request go to that provider, and their own policies apply from there.
- The app does not read your conversations to train anything, and there is no analytics path that carries message text.
If you need help
Local Waifu is not a crisis service and cannot be one. If you are in danger now, contact local emergency services. The banner lists helplines for your language, and the Terms of Use carries the same list.