Privacy & Governance
Trust is not a policy.It is a practice.
The principles below are not a checklist. They are the soil from which everything else grows.
The stories shared in serious illness are among the most sacred a person will ever tell. They deserve more than a privacy policy. They deserve a practice — a discipline of care that governs every decision, every line of code, every model deployment, every governance choice.
What follows is not legal language. It is a sincere articulation of the principles that govern this work — and a future Governance Dashboard that makes every decision visible, auditable, and accountable.
The Four Foundation Principles
Principle 01
Human Oversight, Always
CompassionAI™ is being designed with a human in the loop for every decision that touches a person's care. The AI may listen, may suggest, may notice — but the decisions belong to people.
Principle 02
Privacy as a Vow
The stories shared in serious illness are among the most sacred a person will ever tell. They will never be sold, never used for advertising, never quietly absorbed into a training set without explicit, informed, ongoing consent.
Principle 03
Responsible Artificial Intelligence
Every model is auditable, every decision is explainable, every bias is hunted down and named. We publish what we know and what we do not. We invite the scrutiny of ethicists, clinicians, and the communities we serve.
Principle 04
Future Clinical Connections
CompassionAI™ does not diagnose. It does not prescribe. It does not replace clinical judgement. It is being designed to gently connect people to the right human clinician at the right moment.
Future Prototype · Governance Dashboard
Trust, made visible.
Every AI decision auditable. Every bias hunted. Every governance choice published.
Models in Production
2
Both audited Q3 2025
Ethics Reviews
14
All published
Bias Audits
8
3 corrections made
Consent Withdrawals
100%
Honored within 24h
Recent Governance Decisions
Model deployment paused pending bias audit
Ethics Board
Oct 14
New consent flow introduced for v0.2
Product + Ethics
Oct 02
Warm handoff threshold lowered to 4 minutes
Clinical Advisory
Sep 28
Family Thread feature gated for explicit opt-in
Privacy Officer
Sep 15
Responsible AI Principles
Human oversight on every clinical-adjacent decision
Explainability for every model output
No PHI ever absorbed into training without explicit consent
Bias audits before every model deployment
Consent withdrawal honored within 24 hours
All governance decisions published openly
Patient, family, and clinician voice in every design choice
Audit Trail
Every model decision is logged, reviewable, and transparent.
What we will never do
Lines we will not cross.
- We will never collect protected health information without explicit, informed, ongoing consent.
- We will never sell data. Ever. To anyone. For any reason.
- We will never use patient stories for advertising.
- We will never deploy a model that has not passed a bias audit.
- We will never make a clinical decision. That belongs to clinicians, always.
- We will never operate without human oversight on any decision touching a person's care.
- We will never quietly change a governance practice. Every change will be published.
- We will never stop inviting scrutiny from ethicists, clinicians, and the communities we serve.