ASF Standards · Telemedicine · Standard 10

Standard 10 — Artificial Intelligence Systems for Care

5 criteria · 2 core · 3 standard-level · Version 1.0 · Aligned to ISQua EEA Principle 7, 6th Edition

Platform reliability, security, and data privacy are governed by Standards 1 and 6. This standard addresses the additional governance layer specific to artificial intelligence tools used within the telemedicine service.

Criteria in this standard

10.1 — The Service Never Disadvantages a Patient Who Cannot Access It Digitally
10.2 — AI Systems Are Introduced in Line With Law and Best-Practice Guidance
10.3 — AI Systems Are Genuinely Monitored for Unintended Consequences
10.4 — Clinicians Are Genuinely Consulted Before an AI System Is Introduced
10.5 — Accountability for AI-Assisted Care Is Explicitly Defined
10.6 — Patients Are Genuinely Informed When Their Care Involves AI

10.1

The Service Never Disadvantages a Patient Who Cannot Access It Digitally

Core

A genuine process exists so that a patient who cannot access the telemedicine platform — no reliable internet, no suitable device, limited digital literacy — is never simply left without care, but is genuinely referred or redirected to an appropriate alternative, not told the service is unavailable to them with no real next step.

In plain terms: A patient who can’t actually use the platform still gets genuinely connected to care somehow — not just turned away with nothing offered instead.

Facility category	Crisis	Transition	Small	Standard
Applicability	Adapted	Full	Full	Full

Why this matters

This is marked Core because telemedicine’s entire service model depends on digital access, meaning a patient who genuinely cannot access the platform isn’t facing a minor inconvenience — they face a real, complete barrier to this provider’s care. A genuine, functioning referral or alternative pathway is what prevents digital access itself from becoming the thing that determines whether someone receives care at all.

What good looks like

A genuine, defined pathway exists for patients who cannot access the platform.
Staff are genuinely trained to recognise this specific need.
A real, documented instance shows a patient genuinely redirected.

Common failure modes

A patient who can’t use the platform is simply told the service isn’t for them.

Worked example

In practice

A telemedicine provider with no defined response for digitally excluded patients.

BeforeA patient without reliable internet access called the booking line and was simply told the service required a stable connection, with no alternative offered.

ActionA referral pathway was established to connect such patients with a partner in-person clinic, and staff were trained to offer this proactively.

AfterThe Monitor reviewed a documented referral and confirmed the patient genuinely received appropriate care. Verified.

If you are starting from zero — do this first

Establish a referral relationship with an in-person alternative for digitally excluded patients.
Train staff to proactively offer this, not just state the digital requirement.

The most common mistake: Treating “requires a stable internet connection” as an acceptable final answer with no genuine alternative offered.

Self-assessment questions

1. Is there a genuine, defined pathway for a patient who cannot access the platform digitally? — A real, usable referral or alternative pathway.
Evidence: Referral pathway documentation

2. Are staff genuinely trained to recognise and respond to digital access difficulty? — Real, specific staff awareness.
Evidence: Training record

3. Is there a real, documented instance of a patient genuinely redirected to an appropriate alternative? — A real, concrete example.
Evidence: Referral record

Common reasons for a PARTIAL answer

A referral pathway exists on paper but staff don’t actually know to offer it proactively.

Implementation plan

When	What
Week 1-2	Establish a referral pathway and train staff.

How the Monitor verifies this

Method	What	Detail
DOCUMENT	Referral record review	Reviews a real instance of a patient genuinely redirected.

Supervisor tips

Ask a booking staff member what they’d actually do for a patient with no internet access.

Evidence base

World Health Organization. Global Strategy on Digital Health 2020-2025. Geneva: WHO; 2021.

ASF training courses on GMJ Academy →

Foundation courses A-00 to A-03 are live. Criterion-specific modules are being developed and will link here when published.

10.2

AI Systems Are Introduced in Line With Law and Best-Practice Guidance

Standard

Any artificial intelligence tool used to support triage, diagnosis, or clinical documentation is introduced in accordance with applicable law, or where none exists, with available best-practice guidance such as WHO’s AI ethics and governance guidance — not adopted with no reference to any external standard of responsible use.

In plain terms: Before using an AI tool, the provider has actually checked what the law requires — including across every jurisdiction it actually operates in.

Facility category	Crisis	Transition	Small	Standard
Applicability	Adapted	Full	Full	Full

Why this matters

Telemedicine providers often genuinely operate across multiple jurisdictions, and AI regulation varies meaningfully between them — a governance basis that was genuinely checked for one jurisdiction may not actually apply in another where the provider also delivers care. Real, jurisdiction-specific verification is what this criterion is actually protecting against: an assumed basis that doesn’t genuinely hold everywhere the provider operates.

What good looks like

AI systems are genuinely introduced in line with applicable law.
Where no law exists, introduction is genuinely informed by recognized guidance.
The governance basis is genuinely checked per relevant jurisdiction.

Common failure modes

A single governance basis is assumed to apply uniformly across all jurisdictions served.

Worked example

In practice

A provider operating across three countries with an AI triage tool.

BeforeA single governance review had been conducted for the provider’s home jurisdiction, with no check of whether the same basis genuinely applied in the two other countries served.

ActionA jurisdiction-specific review was conducted for each country, revealing one jurisdiction had a specific AI medical device requirement the provider had not previously met.

AfterThe Monitor reviewed the jurisdiction-specific governance documentation. Verified.

If you are starting from zero — do this first

List every jurisdiction where your telemedicine service is genuinely delivered.
Check AI-specific regulation for each one individually.

The most common mistake: Assuming one jurisdiction’s governance basis automatically applies across every other jurisdiction served.

Self-assessment questions

1. Is any AI system genuinely introduced in line with applicable law? — Real, verified compliance.
Evidence: Regulatory check record

2. Where no law exists, is introduction genuinely informed by recognized guidance? — A real, referenced source.
Evidence: Documented guidance reference

3. Has the governance basis been genuinely checked against each relevant jurisdiction? — A real, jurisdiction-specific check.
Evidence: Multi-jurisdiction review record

Common reasons for a PARTIAL answer

A review covers the provider’s primary jurisdiction only.

Implementation plan

When	What
Week 1-2	Conduct jurisdiction-specific regulatory reviews.

How the Monitor verifies this

Method	What	Detail
DOCUMENT	Multi-jurisdiction review	Reviews the governance basis for each jurisdiction served.

Supervisor tips

Ask about governance in a jurisdiction other than the provider’s home base.

Evidence base

World Health Organization. Ethics and Governance of Artificial Intelligence for Health: Guidance on Large Multi-Modal Models. Geneva: WHO; 2024.

ASF training courses on GMJ Academy →

Foundation courses A-00 to A-03 are live. Criterion-specific modules are being developed and will link here when published.

10.3

AI Systems Are Genuinely Monitored for Unintended Consequences

Standard

The provider genuinely monitors any AI-assisted triage or diagnostic tool for over-diagnosis, missed diagnosis, and other unintended consequences — a genuinely heightened concern in a remote setting where an AI tool may be compensating for the absence of physical examination — not deploying a tool and assuming it works as intended with no real, ongoing scrutiny.

In plain terms: The provider actually keeps checking whether its AI tools work well — this matters even more here, since there’s no physical exam to catch what the AI might miss.

Facility category	Crisis	Transition	Small	Standard
Applicability	Adapted	Full	Full	Full

Why this matters

A remote consultation already lacks the real information a physical examination provides, and an AI tool used to help compensate for this gap carries genuinely elevated stakes — if the tool itself has accuracy problems, there is no physical exam acting as a natural backstop to catch what it misses. Monitoring here needs to genuinely account for this compounded risk, not treat it identically to an AI tool used alongside in-person examination.

What good looks like

AI output is genuinely audited.
Monitoring genuinely reflects remote assessment’s specific limitations.
A documented instance shows monitoring catching a real issue.

Common failure modes

Monitoring treats the remote AI tool identically to an in-person equivalent, missing the compounded risk.

Worked example

In practice

A provider using an AI-assisted symptom triage tool.

BeforeThe tool’s accuracy was audited using the same generic process as an in-house AI tool, without specific attention to cases where the absence of physical examination might compound an AI error.

ActionThe audit was revised to specifically flag and review cases where the AI recommendation diverged significantly from what a subsequent in-person assessment later found.

AfterThe Monitor reviewed the revised audit methodology. Verified.

If you are starting from zero — do this first

Add a specific comparison between AI-assisted remote assessments and any later in-person findings.

The most common mistake: Auditing a remote AI tool the same way as an in-person one, missing the genuinely elevated risk of no physical exam as a backstop.

Self-assessment questions

1. Is AI-assisted output genuinely audited? — Real, ongoing audit.
Evidence: Audit record

2. Does monitoring genuinely reflect remote assessment’s specific limitations? — A real, setting-specific consideration.
Evidence: Audit methodology

3. Is there a documented instance of monitoring identifying a genuine issue? — A real, concrete example.
Evidence: Issue response record

Common reasons for a PARTIAL answer

Audits happen but don’t specifically track divergence from subsequent in-person findings.

Implementation plan

When	What
Week 1-2	Build a remote-specific audit comparing AI output to later in-person findings where available.

How the Monitor verifies this

Method	What	Detail
DOCUMENT	Audit methodology review	Reviews the audit for genuine, remote-specific consideration.

Supervisor tips

Ask whether the audit specifically accounts for the absence of physical examination.

Evidence base

US Food and Drug Administration. Artificial Intelligence and Machine Learning in Software as a Medical Device. Silver Spring: FDA; 2023.

ASF training courses on GMJ Academy →

Foundation courses A-00 to A-03 are live. Criterion-specific modules are being developed and will link here when published.

10.4

Clinicians Are Genuinely Consulted Before an AI System Is Introduced

Standard

Clinicians who will actually use a new AI tool are genuinely consulted before its introduction, with real training needs identified and addressed — not a tool rolled out to a genuinely distributed, often remote clinical workforce with no engagement beyond a brief notification.

In plain terms: Before a new AI tool goes live, clinicians — including those working remotely, not just those at a central office — have had a real say.

Facility category	Crisis	Transition	Small	Standard
Applicability	Adapted	Full	Full	Full

Why this matters

A genuinely distributed clinical workforce is easy to under-consult simply because convening them is logistically harder than gathering an in-person team — but their real, practical experience with the AI tool’s actual use in remote consultations is exactly the knowledge genuine consultation is meant to surface.

What good looks like

Clinicians are genuinely consulted before introduction.
Consultation genuinely reaches remote clinicians across time zones.
Clinicians can describe genuine consultation.

Common failure modes

Consultation happens only with clinicians easiest to convene, missing remote staff.

Worked example

In practice

A provider introducing an AI documentation tool across time zones.

BeforeA single consultation call was scheduled at a time convenient for the provider’s home time zone, which several remote clinicians in other time zones could not attend.

ActionA second session was scheduled, and a written feedback option was added for anyone still unable to attend either.

AfterThe Monitor confirmed clinicians across multiple time zones had genuinely participated. Verified.

If you are starting from zero — do this first

Schedule consultation sessions that genuinely accommodate your clinicians’ real time zones.

The most common mistake: Scheduling a single consultation session at a time that genuinely excludes part of a distributed workforce.

Self-assessment questions

1. Are clinicians genuinely consulted before introduction? — Real, prior consultation.
Evidence: Consultation record

2. Does consultation genuinely reach remote clinicians across time zones? — Real, functioning consultation.
Evidence: Participation record across time zones

3. Can clinicians describe having been genuinely consulted? — Tests whether consultation actually registered.
Evidence: Staff interview

Common reasons for a PARTIAL answer

A single session was held, genuinely excluding part of the workforce.

Implementation plan

When	What
Before any launch	Schedule consultation genuinely accommodating all clinician time zones.

How the Monitor verifies this

Method	What	Detail
ASK	Remote staff interview	Asks a remote clinician whether they felt genuinely consulted.

Supervisor tips

Specifically interview a clinician in a different time zone from the provider’s base.

Evidence base

Topol EJ. High-Performance Medicine: The Convergence of Human and Artificial Intelligence. Nature Medicine. 2019;25(1):44-56.

ASF training courses on GMJ Academy →

Foundation courses A-00 to A-03 are live. Criterion-specific modules are being developed and will link here when published.

10.5

Accountability for AI-Assisted Care Is Explicitly Defined

Core

The provider has genuinely considered and documented accountability for care decisions made with AI support — who is responsible when an AI-assisted decision is wrong — not leaving this as an unexamined question until it actually matters, a question with genuinely added complexity when the clinician, the platform, and the AI vendor may all sit in different jurisdictions.

In plain terms: Everyone knows, in advance, who’s actually responsible if an AI-supported decision is wrong — even with a clinician, a platform, and an AI vendor potentially all in different countries.

Facility category	Crisis	Transition	Small	Standard
Applicability	Adapted	Full	Full	Full

Why this matters

This is marked Core because the cross-border nature of many telemedicine operations adds a genuinely real layer of complexity to accountability that doesn’t exist in a single-site facility — the clinician, the platform operator, and the AI tool’s vendor may each sit in different jurisdictions with different legal frameworks. Without explicit, documented accountability, a safety incident risks becoming a genuinely tangled jurisdictional dispute exactly when the patient needs a clear answer.

What good looks like

Clinical accountability is explicitly documented.
Clinicians genuinely understand accountability, including across jurisdictional lines.
A defined process exists for reviewing accountability in an incident.

Common failure modes

Accountability has never been examined in light of the provider’s genuine cross-border structure.

Worked example

In practice

A provider with clinicians, a platform operator, and an AI vendor in three different countries.

BeforeAccountability for AI-assisted decisions had never been explicitly documented with this genuine cross-border structure in mind.

ActionA clear policy was developed with legal input, explicitly confirming the treating clinician retains clinical accountability regardless of where the AI vendor or platform is based.

AfterThe Monitor reviewed the policy and interviewed a clinician who could clearly describe their own accountability. Verified.

If you are starting from zero — do this first

Map every jurisdiction involved in your AI tool’s use — clinician, platform, vendor.
Document accountability explicitly, with legal input if the structure is genuinely complex.

The most common mistake: Documenting accountability as if the provider operated from a single jurisdiction, missing the genuine cross-border complexity.

Self-assessment questions

1. Is clinical accountability explicitly documented? — A real, clear answer.
Evidence: Documented accountability policy

2. Do clinicians genuinely understand accountability across jurisdictional lines? — Real, demonstrated understanding.
Evidence: Staff interview

3. Is there a defined process for reviewing accountability in a safety incident? — A real, usable process.
Evidence: Incident review protocol

Common reasons for a PARTIAL answer

A policy exists but was never reviewed against the provider’s genuine cross-border structure.

Implementation plan

When	What
Week 1-2	Map jurisdictions and document accountability with legal input.

How the Monitor verifies this

Method	What	Detail
ASK	Staff interview	Asks a clinician to describe their accountability, including cross-border considerations.

Supervisor tips

Ask specifically how jurisdiction affects accountability, not just the general policy.

Evidence base

European Commission. Ethics Guidelines for Trustworthy Artificial Intelligence. Brussels: European Commission; 2019.

ASF training courses on GMJ Academy →

Foundation courses A-00 to A-03 are live. Criterion-specific modules are being developed and will link here when published.

10.6

Patients Are Genuinely Informed When Their Care Involves AI

Standard

A patient whose care involves an aspect delivered with AI system support is genuinely informed of this — not left to assume every decision was made by the remote clinician alone, with disclosure treated as optional rather than a genuine, standard part of informed telemedicine care.

In plain terms: If AI is actually involved in part of a patient’s remote consultation, they’re genuinely told — not left to assume a human clinician made every decision alone.

Facility category	Crisis	Transition	Small	Standard
Applicability	Adapted	Full	Full	Full

Why this matters

A patient’s genuine right to understand their own care includes knowing, in plain terms, when a triage or interpretation decision affecting them was shaped by an AI system rather than the remote clinician’s judgement alone. This matters especially in telemedicine, where a patient genuinely cannot see the clinician’s workflow the way they might observe cues in an in-person visit, making explicit disclosure the only real way they would know.

What good looks like

Patients are genuinely informed when AI is involved in their care.
Disclosure is genuinely understandable, in plain language.
A patient can genuinely confirm they were told.

Common failure modes

AI disclosure is buried in a general platform terms-of-use document, technically present but never genuinely read or understood.

Worked example

In practice

A telemedicine service using an AI-assisted symptom triage tool.

BeforeAI’s role was mentioned only in the platform’s general terms of use, which patients accepted without genuine awareness that an AI tool had actually been involved in their specific triage.

ActionA brief, plain-language note was added during the consultation itself: “An AI-assisted tool helped prioritise your case; your clinician reviewed it directly.”

AfterThe Monitor interviewed a patient who could genuinely confirm they understood AI had been involved. Verified.

If you are starting from zero — do this first

Add a brief, plain-language disclosure point at the moment AI-assisted care is actually discussed with the patient during the remote consultation.

The most common mistake: Treating a line buried in platform terms of use as genuine disclosure, when the patient never actually registers it.

Self-assessment questions

1. Is a patient genuinely informed when AI is involved in their care? — Real, standard disclosure.
Evidence: Disclosure protocol

2. Is this disclosure genuinely understandable, not buried in platform terms? — A real, plain-language explanation.
Evidence: Disclosure wording sample

3. Can a patient asked directly confirm they were genuinely told? — A real, concrete confirmation.
Evidence: Patient interview

Common reasons for a PARTIAL answer

Disclosure exists in platform terms but patients genuinely cannot recall or explain it when asked directly.

Implementation plan

When	What
Week 1-2	Draft plain-language disclosure wording for AI-assisted consultation points and train clinicians to deliver it.

How the Monitor verifies this

Method	What	Detail
ASK	Patient interview	Asks a patient whether they were genuinely told AI was involved in their care.

Supervisor tips

Ask a patient directly rather than relying on the platform terms’ existence alone.

Evidence base

World Health Organization. Ethics and Governance of Artificial Intelligence for Health: Guidance on Large Multi-Modal Models. Geneva: WHO; 2024.

ASF training courses on GMJ Academy →

Foundation courses A-00 to A-03 are live. Criterion-specific modules are being developed and will link here when published.

← Telemedicine Standards — all standards
Self-assess against these criteria →
← Standard 9 — Sustainable Care
Standard 11 — Supporting the Care Workforce →