ANSI National Accreditation Board

The ANSI National Accreditation Board (ANAB) provides comments in the attachment related to Question 16 regarding the involvement of independent third parties as well as limitations cited in the literature regarding a competence-based evaluation of GenAI-enabled devices inspired by the evaluation and credentialing of human clinicians.

The ANSI National Accreditation Board (ANAB) offers the following comments on the Food and Drug
Administration’s August 2026 Considerations for the Regulation of Generative AI-Enabled Medical
Devices: Discussion Paper and Request for Feedback. Specifically, these comments deal with Question
16 regarding the involvement of independent third parties as well as limitations cited in the literature
regarding a competence-based evaluation of GenAI-enabled devices inspired by the evaluation and
credentialing of human clinicians.

The testing and approval of medical devices for fulfillment of regulatory requirements and the
evaluation and credentialing of human clinicians are both examples of the broader concept of
conformity assessment, which is defined by ISO/IEC 17000:2020 as “demonstration that specified
requirements are fulfilled.” Conformity assessment is performed not only to provide assurance that
specified requirements are fulfilled but also to provide assurance that the expected benefits from
fulfilled requirements are realized (e.g., medical devices are safe and effective for their intended use).

In conformity assessment’s most straightforward forms, the object whose fulfillment of specified
requirements is demonstrated is the same object about which assurance is needed. Conformity
assessment for medical devices and conformity assessment for clinicians both follow this pattern.
Medical devices are tested and approved (their fulfillment of specified requirements is demonstrated)
which provides assurance about their safety and effectiveness. Clinicians are evaluated and
credentialed (their fulfillment of specified requirements is demonstrated) which provides assurance
about their competence.

However, under certain circumstances it can be more effective if the object whose fulfillment of
specified requirements is demonstrated impacts but is not exactly the same as the object about which
assurance is intended. An example is conformity assessment for organic food where the objects which
fulfill specified requirements are, for example, the land, processes and products used to grow the food.
When fulfillment of those requirements is demonstrated the assurance intended is not only about those
specific objects but rather also about the resultant food that is grown1.

Such an approach to considering conformity assessment in the regulation of generative AI-enabled
medical devices provides feedback on Question 16 regarding the potential role of third parties and
addresses the limitations cited in the literature regarding a competence-based evaluation of GenAIenabled devices inspired by the evaluation and credentialing of human clinicians.

Specifically, Freyer et al.2 characterize an approach to regulating GenAI applications in a manner
analogous to human clinicians as compelling. In effect, this is the proven approach to conformity
assessment where an object (GenAI application) whose fulfillment of specified requirements is

1 “About the Organic Standards”, USDA Agriculture Marketing Service, accessed October 9, 2026,

https://www.ams.usda.gov/grades-standards/organic-standards
2 Freyer O, Jayabalan S, et al. Overcoming Regulatory Barriers to the Implementation of AI Agents in

Healthcare. Nature Medicine. 2025;31(10):3239-3243.

www.anab.org | Fort Wayne, IN | Washington, DC
demonstrated impacts but is not exactly the same as the object about which assurance is intended (the
medical device).

However, in their representation of possible adaptations to current regulatory frameworks along the
axes of Impact and Feasibility, they present the approach of a human-like regulatory framework as high
impact but low feasibility. They raise the question of accountability of both trainers and trainees in such
an approach. For human clinicians there are professional, legal, and reputational consequences if the
demonstration of fulfillment of competence requirements by the trainee is insufficient in any way or,
after a successful demonstration, misconduct occurs. These consequences are reasonable and realistic
mechanisms to assure accountability regarding human clinician competence for both the trainers and
the trainees/credentialed clinicians. It is noted that such accountability mechanisms would be needed
for demonstrating the competence of GenAI applications in a manner analogous to human clinicians.
Presumably the lack of reasonable and realistic accountability mechanisms is the reason this approach is
reflected as low in feasibility. Several elements of a possible accountability mechanism are suggested
with one element being certification schemes of the type used for medical devices.

ANAB believes the suggestion of a certification scheme is a step toward a reasonable and realistic
mechanism to assure accountability in an approach to regulating GenAI applications in a manner
analogous to human clinicians. However, this suggestion is not consistent with the idea that the object
(GenAI application) whose fulfillment of specified requirements is demonstrated impacts but is not
exactly the same as the object about which assurance is intended (the medical device). To be
consistent, the mechanism for accountability would be certification of the competence of GenAI
applications in a manner analogous to human clinicians as opposed to certification of (or certification
analogous to) a medical device.

When this consistency is followed an obvious answer to Question 16 regarding the role of third parties
also becomes evident.

Certification of the competence of people is an established and proven third-party conformity
assessment activity that provides reasonable and realistic assurance of the competence of an individual.
Thus, a third-party role in a regulatory framework for GenAI-enabled medical devices analogous to
human clinicians would be certification of the “competence” of the GenAI application in a manner
analogous to demonstrating human clinicians fulfill competence requirements.

Such certification for GenAI applications in a manner analogous to credentialing of human clinicians
would introduce comparable accountability. Through either regulatory requirement or market
demands, the attainment of certification becomes an asset to the supplier of the GenAI-enabled medical
device. To protect that asset, the supplier is incentivized to ensure the GenAI application continues to
fulfill requirements. Depending on how the demand or expectation for certification is invoked, real and
substantial consequences (contractual, financial, reputational, and potentially others) will result from
loss of certification should a GenAI fail to maintain fulfillment of requirements.

ISO/IEC 17024:2026 (Third edition) provides requirements for the competence, consistency and
impartiality of certification bodies certifying the competence of people. These requirements could
mostly directly apply to, or in a few cases be easily adapted to certification bodies certifying the
competence of GenAI applications. Such certification in accordance with requirements based on ISO/IEC
17024 would constitute certification for GenAI applications in a manner analogous to credentialing of
human clinicians and establish the needed reasonable and realistic accountability mechanism thus
making this approach of high impact and much more likely feasibility.

The impartiality and consistency related requirements for certification bodies would prevent limiting
competition among GenAI application providers; all providers would have access to a consistent
certification process. The competence requirements for certification bodies would assure that
certification provides trustworthy assurance about GenAI-enabled medical devices.

Further, ISO/IEC 17024:2026 has specific requirements for the process of developing the certification
requirements people must fulfill as well as the examination methods to cost-effectively demonstrate
their fulfillment of requirements. These requirements would likely need some reconsideration to be
most appropriate for competence of a GenAI application. However, such processes could be used for
either or both Device Benchmarking or Clinical Confirmation approaches.

Finally, ISO/IEC 17024:2026 requires that recertification be implemented. Activities in recertification
could be a defined part of overall Postmarket Monitoring.

The content of ISO/IEC 17024:2026 is written for certification as a third-party activity. ANAB offers
accreditation to provide assurance of the competence, consistency, and impartiality of certification
bodies for people via a demonstration that the certification body fulfills the requirements in ISO/IEC
17024. The requirements, conformity assessment, and accreditation infrastructure that exists today for
the certification of people can be adapted to evaluate GenAI applications as a source of assurance for
GenAI-enabled medical devices with reasonable and realistic accountability for the trainers of the GenAI
applications, the performance of the application itself, and the certification bodies who demonstrate the
competence of the GenAI application in a manner analogous to human clinicians. ANAB as the
accreditation body is held accountable via peer assessments from other accreditation bodies for
fulfillment of international requirements for accreditation bodies.

For further information or discussion on these topics, please contact:

Brad Moore
Associate Director Engagement and Strategy
ANAB
bmoore@anab.org

or

Keith Mowry
Associate Director Engagement and Strategy
ANAB
kmowry@anab.org