The system reasons. The clinician decides.
The Advisory Principle is how we govern every AI surface in HealthOS. AI outputs are advisory, not declarative. Clinicians and institutional operators are the authority. The governance framework is architectural — not contractual.
Last reviewed:
Advisory Principle
Clinical Advisory Board review
Model versioning
Public change log
Clinician override
Audit trail
Ethics Council oversight
The Advisory Principle.
Every AI output in a care-affecting context is advisory. The clinician decides. Every override is recorded with a reason. The system does not act autonomously on clinical decisions.
Clinical Advisory Board review.
Every model with clinical effect is reviewed before production release. Reviews evaluate training data, evaluation metrics, known limitations, failure modes, and clinical applicability.
Model versioning.
Every production model carries a semver, a deployment date, a training snapshot reference, and an evaluation record. Changes are logged in the public model change log.
Dated change log.
The model change log at /trust/responsible-ai lists every production model version with deployment date, training reference, evaluation summary, and Clinical Advisory Board review.
Clinician override.
Every AI recommendation is overridable as a single explicit action. The override is recorded with reason and contributes to supervised retraining.
Audit trail.
Every inference carries a reasoning trace. Every use of an AI output is auditable by clinicians, institutional operators, regulators, and ethics councils.
Ethics Council.
Veronara's ethics council — named on /trust/ethics-council — provides institutional oversight over clinical AI decisions, model retraining cadence, and ethical edge cases.
Frequently asked
Common questions about this layer.
What is the Advisory Principle?
The Advisory Principle is Veronara's institutional commitment that AI in HealthOS issues advisories — not decisions. The clinician is the decision-maker; the model is the advisor. Every advisory is logged, attributable to a specific model version, and subject to clinical review.
How is model governance handled?
Through institutional process — pre-deployment safety assessment, post-deployment monitoring, drift triggers, and retirement criteria. Quarterly review by the Clinical Advisory Board with the responsible clinical lead, the model owner, and external clinical input as appropriate.
What happens when a model drifts?
Drift triggers are defined per prediction before deployment. On a trigger: clinical leadership is notified, institutional review is initiated, and the model may be flagged as 'under review' in clinician workflows until review concludes. A retired model is documented and dated; the audit trail is preserved.
Are AI safety incidents disclosed publicly?
Material operational incidents — including those affecting clinical AI safety — are disclosed publicly within 72 hours at /trust/incidents, with root cause, remediation, and data impact, preserved without silent edit.
Who reviews AI safety at Veronara?
The Clinical Advisory Board — seven seats with named mandates including a Clinical Ethics Lead, the Chief Medical Officer, and specialty leads. The board is structurally independent of executive leadership; at least four of seven seats are filled by clinicians external to the company.
Safety is not a feature. It is the architectural commitment under which every clinical workflow operates. The Advisory Principle is non-negotiable.
Signed by the Veronara Clinical Safety Office · Last reviewed · Propose a correction to corrections@veronara.com.
Engage Veronara.
Executive briefings are offered to hospital networks, ministries of health, and enterprise healthcare institutions.
For hospital networks and enterprise healthcare institutions.
Acknowledged within two business days.
For ministries of health, national digital health programs, and sovereign deployments.
Acknowledged within 72 hours.