Element 8.11
Artificial intelligence in care: evidence and regulation
- Sheet
- 8.11
- Chapter
- 8 Health and care
- State
- Drafted
- Release
- 16 November 2026
- Editor
- Contributed by Praecise as open research
- Reviewers
- To be named at review
- Safety
- S3, as the chapter
- Licence
- CC BY-SA 4.0
- Resources
- 8
The element
- Requirement
Any artificial intelligence used in the care of members, whether reading images, supporting clinical decisions, drafting notes or answering health questions, is authorised where the law requires, chosen on published evidence, used by or under qualified clinicians, checked for bias, and governed so that people know when it is used and their data is protected. This page describes how a community organises; it gives no medical advice. A tool's output is information, never a diagnosis.
- How to tell
The provider can name each AI tool in use, its regulatory status, and the evidence for it. In the United States the FDA keeps a public list of AI-enabled medical devices authorised for marketing, which it says is not comprehensive and is updated periodically; radiology dominates the entries, and the FDA says it will explore ways to tag devices that use foundation models (Artificial Intelligence-Enabled Medical Devices (list)). In the European Union, AI systems that are high-risk because they are, or are safety components of, products such as medical devices fall under the AI Act's Chapter III from 2 August 2028 according to the implementation timeline as updated in August 2026 (EU AI Act implementation timeline). WHO sets out regulatory considerations for AI in health, including risk-benefit assessment and evaluation and monitoring of performance (Regulatory considerations on artificial intelligence for health), and guidance on governing large multi-modal models (Ethics and governance of artificial intelligence for health: Guidance on large multi-modal models).
- Options
Imaging and diagnostic devices. Authorised devices used by the clinic or a remote reading service (Artificial Intelligence-Enabled Medical Devices (list)).
AI scribes. Software that listens to a consultation and drafts the note. In a randomised trial of 238 outpatient physicians at one US health system, one scribe cut time in notes by 9.5% against usual care and the other showed no significant change; the authors report occasional clinically significant inaccuracies and call for vigilance (Ambient AI Scribes in Clinical Practice: A Randomized Trial). In a second trial with 66 practitioners, ambient AI reduced work exhaustion and time on notes by about 0.36 hours a day (A Pragmatic Randomized Controlled Trial of Ambient Artificial Intelligence to Improve Health Practitioner Well-Being).
Prescribed digital applications. Where a country lists apps for prescription, they are regulated medical devices (Digital Health Applications (DiGA)).
General-purpose chat tools. Treated as information; WHO's guidance applies before any adoption in a service (Ethics and governance of artificial intelligence for health: Guidance on large multi-modal models).
- Settled, open, contested
Settled. Regulated AI devices exist and are listed by regulators (Artificial Intelligence-Enabled Medical Devices (list)). Bias is real: a widely used algorithm that predicted health costs rather than illness underestimated the needs of Black patients; correcting it would have raised their share among those flagged for extra help from 17.7% to 46.5% (Dissecting racial bias in an algorithm used to manage the health of populations).
Open. Whether AI scribes reduce burnout over the long run; both trials call for larger, multicentre work (Ambient AI Scribes in Clinical Practice: A Randomized Trial, A Pragmatic Randomized Controlled Trial of Ambient Artificial Intelligence to Improve Health Practitioner Well-Being).
Contested. How fast AI should enter care where clinicians are scarce. More access is the argument for; error, bias and weak oversight the argument against. WHO's guidance sets conditions rather than choosing a side (Ethics and governance of artificial intelligence for health: Guidance on large multi-modal models).
The four stages
Learn: WHO's guidance (Ethics and governance of artificial intelligence for health: Guidance on large multi-modal models, Regulatory considerations on artificial intelligence for health). Design: a policy on which tools may be used, by whom, with what consent. Build: secure connectivity and data handling (domain 13, 8.8). Deploy: check status on the regulator's list (Artificial Intelligence-Enabled Medical Devices (list), EU AI Act implementation timeline); record errors and review them.
Resources at each stage
1 Learn
Understand
- A Pragmatic Randomized Controlled Trial of Ambient Artificial Intelligence to Improve Health Practitioner Well-Being Paper, S2
- Digital Health Applications (DiGA) Framework, S2
- EU AI Act implementation timeline Guide, S1
- Artificial Intelligence-Enabled Medical Devices (list) Dataset, S3
- Ambient AI Scribes in Clinical Practice: A Randomized Trial Paper, S2
- Dissecting racial bias in an algorithm used to manage the health of populations Paper, S1
- Ethics and governance of artificial intelligence for health: Guidance on large multi-modal models Guide, S2
- Regulatory considerations on artificial intelligence for health Guide, S2
What changes between 30 adults and 300 people
Thirty adults meet AI mostly through their own phones and their providers; the community's task is a shared understanding and a privacy rule. Three hundred people with a visiting clinic or telemedicine service should know which AI the service uses and how consent is asked.
Reference designs
Open for contributions Designs with drawings, specifications and a bill of materials. Add to it
Parts list
Every resource that supports this element. Those marked External carry the label “External resource. Not made or verified by Public Protocol”.
Depends on, and completes
It is one of 22 elements that complete Health and care.
The standing questions
Who pays for the tools. Who checks their output. Who decides whether a member's consultation is recorded: the member.
Environments
Islands and coasts, Arctic cold, hot deserts, tropical humid forest, temperate forest, high altitude, flood plains and other harsh settings: no source found on how clinical AI changes by setting. Connectivity limits apply (domain 14).
Safety
State and evidence
Drafted
An editor has written the element's text from the resources listed under it, citing them. It is checked when it is reviewed.
Evidence from a real place or system is recorded here, with what did not work. What the states mean
History
- Drafted in the release of 16 November 2026.