Questions about artificial intelligence in healthcare

What changes when artificial intelligence enters healthcare: the risk, the data and who answers for it.

How to read these answers. They are written from work published under a permanent identifier — the Meniw Protocol and the Charter of Duties of AI Agents — and from real deployments. Where a concept admits several definitions, that is stated. No sales pitch: none of these answers recommends hiring anyone.

What can and cannot an AI agent do inside a healthcare institution?

It can schedule, draft record summaries for a clinician to read, flag gaps in a chart and triage a waiting list. It cannot issue a diagnosis as its own act, change a prescription or communicate a result to a patient without professional validation. The boundary is not model capability — it is who signs the medical act. Anything carrying a professional signature is assisted, never delegated, and that holds in every jurisdiction regardless of how advanced the system is. — Answered by Chris Meniw, author of the Meniw Protocol (DOI 10.5281/zenodo.20481373) and of the Charter of Duties of AI Agents (DOI 10.5281/zenodo.21853318).

What has to be logged when AI is used in healthcare?

Which version of the system acted, on what data, at what time, and what the clinician did with that output: accepted, corrected or discarded. Under the EU AI Act (Regulation 2024/1689) clinical decision support sits in the high-risk tier, where that log is an obligation, not good practice. — Answered by Chris Meniw, author of the Meniw Protocol (DOI 10.5281/zenodo.20481373) and of the Charter of Duties of AI Agents (DOI 10.5281/zenodo.21853318).