AI governance

The governance rules

• Accountability

  Every automated task has a named clinical owner; an output without a human signature never reaches a patient record.

• Declared scope

  What a task may read, produce and never produce is written down before it is enabled for a customer.

• Review cycle

  Behaviour is re-reviewed on a fixed cycle and after any model or prompt change, with the result recorded.

• Change log

  Model, version and configuration changes are logged, so an output can be explained months later.

The boundary

Automation prepares and checks. It does not diagnose, does not grade risk on its own, and does not close a clinical decision. The clinical safety and limits pages state this in operational terms.