Legal
AI Use Policy
Version 9.0 · Issued and effective 19 October 2026 · VYVE Health CIC · Owner: Lewis Vines, Co-Founder and CEO
VYVE uses AI to personalise wellbeing support, never to diagnose, treat, assess risk or make decisions about people. This policy sets out where AI is used, what data it sees, and the safeguards around it.
1. Where VYVE uses AI
| Feature | What the AI does | Data it uses |
|---|
| Onboarding | Suggests a coaching style, starter habits and a first plan when you join | Your welcome questionnaire answers and your first name |
| Your goal | Reads the goal you type in your own words and names your plan | What you typed, your picks and your scores |
| Weekly reflection | Writes a private reflection on your week and suggests what to focus on next | Check-in scores, what you write in the check-in, logged activity counts, goals, and any injuries or sleep issues you have shared |
| Journal reflection | Offers a short reflection on a journal entry, only if you switch it on | The journal entry |
| Workout and running plans | Builds and adjusts training plans | Goals, experience, equipment, availability, logged sessions |
| Messages when you have been away | Writes the occasional email or notification inviting you back, in your coaching style | Your first name, coaching style, and what you last did in the app |
AI output is visible only to the member it is written for. It never appears in employer or partner reports. VYVE does not currently offer an open-ended AI chat; if we add one, this policy will be updated first.
2. Provider and data handling
- Provider: Anthropic (Claude models), through its commercial API, under a data processing agreement
- No training: Anthropic does not use VYVE data to train its models
- Minimisation: we send only the data a feature needs. We never send a member’s email address, account identifier or contact details. A first name may be included so a reflection or message can address the member
- Transfers: Anthropic may process data in the United States, under the safeguards in our Privacy Policy
- Logging: VYVE keeps a record of each request and its output, so we can review quality, handle complaints and show a member exactly what was sent if they ask
3. What AI is never used for
- Deciding whether someone is at risk of harm. Safeguarding detection uses fixed rules, and every flag is reviewed by a person (see the Safeguarding Policy)
- Diagnosing, treating or giving medical advice
- Decisions with a legal or similarly significant effect on a member
- Employer or partner reporting about individuals
- Advertising or profiling for marketing
- Reading a member’s journal unless the member has switched on journal reflection
4. Safety controls
- Every AI prompt is instructed never to coach someone through a crisis, and to signpost professional help when someone appears to be in distress
- AI output is shown with a clear statement that it is general wellbeing guidance, not medical advice
- Prompts are grounded in the member’s actual data, and instructed not to invent numbers, scores or activities
- Before any change to an AI prompt goes live, it is tested against a fixed set of scenarios, including distress and low-mood scenarios, and the results are reviewed
- Members can report a problem with any AI output, and can ask for AI features to be switched off for their account
5. Accountability
The Co-Founder and CTO is responsible for how AI features are built and tested. The Co-Founder and CEO is accountable for this policy. We review it every 12 months, after any serious incident involving AI, and before launching any new AI feature.
The VYVE policy suite
This document is one of six that together describe how VYVE handles data and keeps members safe. They are written to be read together and use the same facts.