AI disclaimer and safety

What it is

A companion that listens, helps you find words, prays a short prayer with you and points you to Scripture that fits. It runs on self-hosted, open-source language models, and every reply follows the same shape: acknowledgement, prayer, Scripture, encouragement, and sometimes a gentle question.

What it will never do

Emergencies

If you mention self-harm, suicide, abuse, violence or a medical emergency, ShortPray AI stops offering guidance and asks you to contact local emergency services or a crisis line, and to tell a person near you. It will still pray with you — but human help comes first, and asking for it is never a failure of faith.

Scripture accuracy

The model is not allowed to write Scripture. Verses are retrieved from a licensed verse database and quoted with the reference and translation. If nothing suitable is found, the assistant says so and offers a theme instead of inventing a citation.

Where it runs

Self-hosted open-source models first (Ollama and llama.cpp), verified free APIs only as a fallback during an outage, and a deterministic in-house responder behind those. No paid AI subscription, no keys in your browser, and no vendor receives your account details.

Report a bad reply

If the assistant says something judgemental, theologically reckless or unsafe, report it from the reply or through the contact page. Reported replies are reviewed by a human and used to tighten the safety rules — never to train a model on your words.