AI Disclaimer
How to read what the model says — a draft, not advice.
A draft, not advice
Tacita’s replies come from a language model running entirely on your phone. The model is good at sounding confident and bad at being right. Treat what it says as a draft you should verify, not as authoritative advice.
It can be wrong
The model has no live access to the internet and no way to verify its claims. It will sometimes invent facts, mis-quote sources, get dates and numbers wrong, and be confidently mistaken about things you can easily check.
It can be biased or offensive
The model was trained on a large body of human writing and reflects the biases of that material. It may produce content that is offensive, stereotyped, or harmful. You are in charge of what you ask it and what you do with the answer.
Not a substitute for a professional
For medical, legal, financial, psychological, or safety-critical questions, consult a qualified human. A model trained months ago, with no idea who you are or what your situation actually is, cannot replace someone who can.
In an emergency
Tacita does not know it is an emergency. If you or someone else is in danger, call your local emergency services. In the European Union: 112.
Inference is local
The developer cannot see, store, or moderate what you write or what the model produces. There is no remote safety net. The privacy guarantee and the lack of remote moderation are two sides of the same coin.
You acknowledged this disclaimer when you set up the app. You can re-read it here at any time.