Art. 01
We cite, then we claim.
When askFinz makes a factual statement in research mode, it must be able to point at the source it came from. If the source isn't available, the claim isn't made. Footnotes are first-class, not decoration.
Trust · the constitution · v1
Nine commitments we hold ourselves to,
in the order we'd defend them.
Art. 02
We don't train on your data by default.
Your prompts, files, and interactions are not used to train our models unless you explicitly opt in. Training is opt-in, never automatic. If you choose to help improve the models, only the data you opted into is ever used — and you can opt out again at any time. You can review our data practices in privacy policy.
Art. 03
We improve with your permission, not without it.
askFinz gets better over time, but never at the quiet expense of your privacy. We only learn from the work you've explicitly opted in to share, and that choice is yours to reverse whenever you like. Sensitive workspaces stay isolated and are kept out of any improvement loop entirely.
Art. 04
We keep humans in the loop, on purpose.
Our agents are not autonomous. The Navigator pauses for CAPTCHAs and unfamiliar pages. The Coder asks before it commits. ML-predicted check-ins surface decisions to you before they harden. We will not be optimising these pauses out of the system.
Art. 05
We disclose what powers each turn.
You can always see which model wrote the answer in front of you. Routing is intent-aware but never invisible — see /models for the providers we route to today.
Art. 06
We isolate what should be isolated.
Sensitive workspaces (Med is the first example) run on their own database with separate credentials. The default is “closer to private than convenient”.
Art. 07
We won't lock you in.
You can export your data. You can bring your own model keys. You can leave. We'd rather lose a customer to a better tool than win a customer who can't.
Art. 08
We change this document in public.
When something here changes, we publish the diff in /changelog and email anyone in private access. We don't silently widen our own permissions.
Art. 09
We tell you who wrote what.
Every answer in askFinz carries a small badge naming the AI behind it. We pick the right one for the task, but we never hide the choice. You can always trace a paragraph back to its source — the writing, and the citations.
How this document changes
We treat this page like a constitution, not a marketing flyer. When an article tightens, broadens, or is added, the diff is published at /changelog and emailed to everyone in private access. We don't edit articles silently; if you believed something on a Tuesday, you'll see it move on a Wednesday.
Disagree with one of these? We'd like to hear it. Email trust@askfinz.com.