Implemented policy

AI safety and human review

Typed capabilities, bounded delegated context, source citations, uncertainty, and human decision gates make AI assistance inspectable.

Product evidenceImplemented policy

AI safety policy

Capability · scope · uncertainty

Draft → review → clinician decision

Source-linked product view · patient-safe evidence · governed action boundary

Product evidence preview for AI safety and human review; open the linked workspace surface for the live product route.

How AI improves this feature

The safety layer is the feature: models may organize, draft, and flag; they may not silently write the clinical record or take consequential action.

Open the product surface

Explore the current workspace route and compare the implemented evidence with the public explanation.