Responsible AI Statement | Epinova Health

AI can help public health teams see risks earlier and act more precisely - but only if it is trustworthy. These principles guide how we build Epinova. They will be refined with partners, advisors and affected communities.

1. Human oversight

Epinova provides decision support. Qualified people review forecasts, rankings and recommendations and remain accountable for decisions. The platform is not designed to take autonomous actions affecting people.

2. Source traceability

Alerts, forecasts and assistant answers should be traceable to the data, sources and model versions that produced them.

3. Validation before use

Models are back-tested on held-out data and evaluated against simple baselines before operational use. We report performance, including where models perform poorly, and monitor for drift after deployment.

4. Honest uncertainty

Forecasts are presented with prediction intervals and data-quality indicators. We avoid language and visuals that imply false precision.

5. Equity

Data gaps often affect the most marginalized communities. We look explicitly for under-reporting and bias, and design prioritization to make equity trade-offs visible rather than hidden.

6. Privacy and data minimization

We work with aggregate, community-level data wherever possible, collect only what is needed, and protect any personal data with strong safeguards.

7. Data sovereignty

Partners and countries retain ownership of their data. Data is used only for agreed purposes under formal agreements.

8. Generative AI safeguards

Language models are used for tasks such as translation, extraction and grounded question-answering. Outputs are grounded in authorized sources, cited, and designed to decline when evidence is insufficient. Generative features are evaluated before any release.

9. Transparency about status

We state clearly which capabilities exist, which are in development and which are planned, and we label all demonstration data as simulated.