
Responsible AI
Grove uses AI to save educators the second shift, not to judge children. This is what the model does, what it is never allowed to do, and who is accountable for what a family reads.
On this page
What the AI actually does
An educator captures a moment in the field: a photograph, a spoken sentence, a tag or two. The AI turns that capture into a draft observation in plain language and suggests which curriculum strands it touches. That is the whole job.
What it is never allowed to do
- No scores, grades, levels or rankings of children.
- No comparison of one child against another or against a norm.
- No prediction of behaviour, risk or future performance.
- No facial recognition or biometric identification.
- No publishing to a family without an educator's approval.
A person is always in the loop
Nothing reaches a family automatically. The educator who was there edits the draft, corrects it, or throws it away. An observation carries their name because the judgement is theirs.
What the model is shown
Only the capture being drafted and its tags. The model is not given a child's history, other children's records, health notes, or anything about a family. Names are minimised in prompts wherever the draft does not need them.
Your data is not training data
No school's content is used to train foundation models, ours or anyone else's. Our AI providers are contractually barred from retaining prompts or outputs beyond the processing of the request.
Where the AI runs
Drafting runs on named commercial models under enterprise terms, in Canadian regions where the provider offers them and in the United States otherwise. Where a capture can be handled on the device — transcription, for instance — we prefer that.
When it gets things wrong
It will. A draft may misread what a photograph shows or reach for a curriculum link that is not there. That is why the educator approves it, and why every observation keeps an edit history: a family can see that a draft was written by Grove and approved by a named educator, and the school can see what changed.
Language and bias
We write to a plain-language style guide: describe what a child did, never what kind of child they are. Drafts avoid deficit framing, diagnostic vocabulary and comparison. We review samples of live drafting for language that slips, and we correct the prompts rather than asking educators to work around them.
Labelled, always
Anything the AI drafted is labelled as AI-drafted and educator-approved where the family reads it. We do not present machine writing as if a person wrote it unaided.
Turning it off
A school can switch AI drafting off for the whole school, and a family can ask their school to switch it off for their child. Capture, attendance and the record all keep working; educators simply write the observations themselves.
Oversight
We review drafting quality and any reported AI incidents quarterly, and we tell schools when a change to the model or the prompts changes what their staff will see. Concerns can be raised at any time through the contact form.
