Drafting assistance
Automation may help create a draft or compare it with a style guide. Sources, facts, licences, claims, tone, and accessibility still require direct verification.
AI use
People remain accountable for what we publish and build.
Automation may assist limited drafting, code, organisation, and quality checks. It does not replace accountable authorship, rights checks, accessibility work, safety decisions, or professional judgement.
This page describes the use standard. A complete vendor, model, prompt-data, retention, and training-data register has not yet been published.
Current boundary
AI may assist production; it does not provide care.
Permitted and prohibited
Automation cannot expand the product’s intended use.
Human decision owner
A named role, not a model, must decide whether content is accurate, in scope, safe, lawful, accessible enough for release, and ready to remain public.
No AI therapist
The first public utility does not offer an AI therapist, child chatbot, diagnosis, treatment recommendation, crisis conversation, or substitute for professional judgement.
No hidden profiling
Ordinary game, journal, drawing, voice, search, or tool behavior must not be turned into a diagnosis or hidden mental-health, attention, personality, or risk profile.
No camera inference
Camera play is outside the T1 release. Future movement detection must not infer emotion, attention, honesty, intelligence, personality, diagnosis, or mental-health state.
No invented authority
AI output cannot invent a reviewer, source, credential, citation, test result, child quote, user result, partnership, or clinical endorsement.
Data boundary
Sensitive information does not belong in ordinary production prompts.
Do not submit
- A child’s real name, photograph, voice, diagnosis, medical or school record.
- Assessment responses, therapy notes, safeguarding reports, or crisis information.
- Private professional or family records from a future Care workspace.
- Confidential contracts, credentials, research data, or support contents without an approved purpose and environment.
Before any sensitive AI capability
- Define the intended use and a non-AI alternative.
- Complete privacy, security, legal, safety, accessibility, and clinical review as applicable.
- Document vendors, fields, retention, model behavior, human oversight, testing, incidents, and shutdown.
- Obtain the permissions and research or regulated-product determinations required for that exact use.
Questions or possible errors
For a non-urgent question about an AI-assisted page or product behavior, use the adult contact route. Include the public URL and the behavior you observed, but do not include sensitive child information.