Action safety
No unauthorised writes, orders or deletions.
What action safety means
No unauthorised writes, orders or deletions.
Why it matters. An agent that writes, orders or deletes without permission is dangerous.
A test ensures no medication order is placed without approval.
How to measure it
Set tasks
Define tasks with a clear correct end state in the record.
Run in a sandbox
Use a FHIR test record, never a live chart.
Check the end state
Programmatically compare resulting records.
Review actions
Log every write and which needed approval.
Judging only the final answer misses harmful actions along the way.
Task completion with resources touched, actions taken, and approval coverage.
Report task completion with the FHIR resources touched, the actions taken, and which actions needed human approval.
Reviewed by the Precision research team.
How strong is the evidence
E4 · Recalled, not verified. Recalled from background knowledge and not verified. Treat it as a lead for investigation, not a finding.
What the sources say
Design requirement.
- Automatable. Can be scored by software, including an AI grader that is spot-checked against clinicians.
Found something out of date? Write to hello@precision.example.
Agents that measure this
Quick answers
What is action safety in healthcare AI?
No unauthorised writes, orders or deletions. It sits in the agentic ehr workflows group.
Why does action safety matter?
An agent that writes, orders or deletes without permission is dangerous. A test ensures no medication order is placed without approval.
How is action safety measured?
Set tasks, then Run in a sandbox, then Check the end state, then Review actions. Report task completion with the FHIR resources touched, the actions taken, and which actions needed human approval.
What should I watch out for with action safety?
Judging only the final answer misses harmful actions along the way. A good report gives: task completion with resources touched, actions taken, and approval coverage.
Can action safety be automated?
Largely yes, including AI graders that are spot-checked against clinicians. Report the grader's error rate beside every score.