Standards › Standard 2D › Policies
Policy · AI-POL-002 · Version 1.1 · Approved by the ASF Expert Committee
Supports: 2D.7
At a glance
- The clinician who makes the final decision remains accountable for it.
- Write the accountability arrangement for each AI system, approved by the governing body.
- Review accountability after any AI-related safety incident.
- Make the arrangement known to every clinical user.
- Stop and get help straight away if: an AI-assisted decision is made where accountability is not written down for that system.
1. Purpose
To define, in writing and in advance, who is accountable when an AI-assisted care decision is wrong, so that the question is answered before it matters [E-011, EC].
2. Scope
| Applies to | Clinicians, clinical leads, chief executive, governing body, legal adviser |
| Locations | All clinical areas using AI systems at the facility |
| When | Before each AI system is introduced; reviewed yearly and after incidents |
| Does not apply to | Accountability for decisions made without AI, covered by existing clinical governance |
3. Definitions
| Term | Meaning |
|---|---|
| Accountability | The duty to answer for a decision, and to put right harm caused by it |
| Final decision | The clinical decision that is communicated to the patient and recorded |
4. Background
When an AI-supported decision goes wrong, the question “who is responsible” is often left open, which harms patients and staff. Criterion 2D.7 asks for it to be defined in advance and approved [E-011].
5. Responsibilities
| Role | Responsibility | Competence |
|---|---|---|
| Governing body | Approves the arrangement | Governance |
| Chief executive | Keeps the arrangement current; leads reviews | Executive |
| Clinicians | Accept accountability for their final decision and understand the system limits | Clinical |
| Legal adviser | Advises on the legal position (EC) | Legal |
6. Materials and equipment
| Item | Requirement | Condition |
|---|---|---|
| Accountability arrangement per AI system | Written, approved by governing body | Current |
| Staff briefing on accountability | Plain language | Given at induction and when a system is introduced |
7. Safety
- The arrangement must never discourage clinicians from overriding an AI output.
8. Before you start
- Confirm the arrangement for the system is approved.
9. Procedure
Step 1. Write the arrangement.
- For each AI system, write who makes the final decision, who is accountable for it, and how the system’s limits are communicated.
- Include the process for patient harm and redress.
Key point: Write it before use, not after the first incident.
Step 2. Approve.
- The governing body approves the arrangement before the system is used.
Key point: Approval makes the arrangement binding on leaders as well as staff.
Step 3. Brief clinicians.
- Brief each clinical user on their accountability and on the limits of the system.
- Record the briefing.
Key point: Clinicians must understand their duty before they use the tool.
Step 4. Review after incidents.
- After any AI-related incident, review whether the arrangement was clear and worked.
- Update it and record the change.
Key point: Each incident tests the arrangement.
Step 5. Review yearly.
- Review each arrangement yearly, and after any change of system or law.
Key point: Law and systems change.
10. Decision points
- If accountability is unclear in an incident, the chief executive decides and records the reasons, then updates the arrangement.
11. Special situations
| Situation | Action |
|---|---|
| Clinician overrides an AI output and harm follows | Review the decision under the incident process, not automatically as an AI fault. |
12. Escalation
- AI-related incidents go to the chief executive and INC-POL-001 at once.
13. Monitoring
| Measure | Target | Frequency | Owner | If not met |
|---|---|---|---|---|
| AI systems with approved arrangement | 100% | Yearly | Chief executive | Stop use until approved |
| Clinicians briefed | 100% | Yearly | Clinical leads | Brief within 30 days |
| Incidents reviewed against arrangement | 100% | Each event | Chief executive | Escalate to governing body |
14. Records
| Record | Kept in | Retention |
|---|---|---|
| AI accountability arrangement and briefing records | Governance file | the period set by the facility’s records policy years |
15. Training
- Clinicians are briefed at induction and when a system is introduced (EC).
16. Related documents
| ID | Title | Use |
|---|---|---|
| AI-POL-001 | Responsible Use of AI in Care Policy | Parent policy |
| INC-POL-001 | Incident Reporting and Management | Incident route |
| INC-SOP-001 | Root Cause Analysis of Serious Incidents | Review of serious incidents |
17. Evidence
| Reference | Note |
|---|---|
| E-011 | WHO (2021), Ethics and governance of artificial intelligence for health. Used as the basis for AI governance. Link and wording to verify before publication. |
| E-009 | ASF Expert Committee consensus. Legal position set by the ASF Expert Committee. |
18. Change history
| Version | Date | What changed | Approval |
|---|---|---|---|
| 1.0 | the date of approval | First version | Approved by the ASF Expert Committee |
| 1.1 | the date of approval | Internal quality review: corrected cross-references, status wording and document control | Approved by the ASF Expert Committee |
E-learning for this document
References
- World Health Organization (2021). Ethics and governance of artificial intelligence for health: WHO guidance. Geneva: WHO. Open source (E-011)
- Accréditation Sans Frontières (2026). ASF Expert Committee consensus on Standard 1. (E-009)