This document is a good-faith draft written to describe what OnBetterTerms actually does. It has not been reviewed by a lawyer and is not jurisdiction-specific. It must not be relied upon as a binding agreement until counsel has reviewed it. See docs/LEGAL-REVIEW.md for the outstanding items.
AI Disclosure
Version 2025-01-01
You are talking to software
Guidance in OnBetterTerms is generated by large language models. It is not a person, not a licensed professional, and not supervised in real time.
It can be wrong
Language models can produce confident, plausible, incorrect statements. Guidance here is a prompt for your own thinking, not a conclusion to act on without judgement.
What is sent to model providers
Content is classified before any model call. Personal sessions may include your own private content; shared sessions are built from a separate path that can only read shared content. Content marked 100% private never enters any request whose output another member can see.
Model routing
Different workflows use different models chosen for the task. Safety assessment uses a high-capability model and is never billed to you.
Longitudinal observations
Insights describe correlations in your own self-reported data. They are not causal claims and not clinical assessments. We do not produce probability estimates about whether a relationship will succeed or end.
Your controls
You can rate any response, view and delete what the AI remembers, disable memory categories, and turn predictive insights off entirely.
Training
Your content is not used to train models.