AdvaMed (Geeta Pamidimukkala, Zack Hornberger)

Please see attached comments.

1301 Pennsylvania Avenue, NW
Suite 400
Washington, D.C. 20004
P :: 202.783.8700
F :: 202.783.8750
W:: AdvaMed.org

October 19, 2026

Dockets Management
Food and Drug Administration
5630 Fishers Lane, Rm 1061
Rockville, MD 20852

Re: Considerations for the Regulation of Generative AI-Enabled Medical Devices:
Discussion Paper and Request for Feedback, Docket Number FDA-2026-N-7874

To Whom It May Concern:

The Advanced Medical Technology Association (AdvaMed) appreciates the opportunity to
submit comments in response to the Considerations for the Regulation of Generative AI-Enabled
Medical Devices: Discussion Paper and Request for Feedback 1 (hereafter referred to as the
discussion paper or the paper).

AdvaMed is the world's largest association representing manufacturers of medical
devices, diagnostic products, and medical technology. AdvaMed's member companies
range from the largest to the smallest medical product innovators and manufacturers.
AdvaMed's member companies produce innovations that transform health care through
earlier disease detection, less invasive procedures, and more effective treatments.
AdvaMed advocates for a legal, regulatory and economic environment that advances
global health care by assuring worldwide patient access to the benefits of medical
technology. The Association promotes policies that foster the highest ethical standards,
rapid product approvals, appropriate reimbursement, and access to international markets.

AdvaMed commends the Food and Drug Administration (FDA or “Agency”) for its publication
of this discussion paper and its commitment to seeking stakeholder feedback in the development
of regulatory considerations for generative AI (GenAI) enabled medical devices. The
development and implementation of clear, effective, and consistent regulatory policies and
expectations for GenAI-enabled medical devices is critical to ensuring patient safety while
fostering innovation.

GenAI-enabled medical devices hold potential to improve patient outcomes, enhance clinical
workflows, and expand access to care. The discussion paper offers a timely opportunity for
stakeholders to contemplate and provide feedback on relevant premarket and postmarket

1 Available online: https://www.fda.gov/medical-devices/digital-health-center-excellence/considerations-regulation-generativeai-enabled-medical-devices-discussion-paper-and-request

advamed.org :: @AdvaMedUpdate :: AdvaMed
FDA Dockets Management Docket No. FDA-2026-N-7874
October 19, 2026
Page 2 of 9

regulatory considerations for GenAI-enabled devices. FDA’s regulatory approach to all medical
devices, including GenAI-enabled devices, should support, rather than hinder, innovation. This
objective can be achieved through a balanced approach; one that prioritizes risk-based
regulation, alignment with existing regulatory frameworks, and flexible postmarket oversight.
We urge the FDA to consider these principles to reinforce its commitment to both patient safety
and technological progress. We respectfully submit for your consideration our feedback and
response to the discussion paper.

Regulatory approaches for GenAI-enabled devices should remain centered around riskbased principles
FDA’s discussion paper notes that GenAI is a technology that may introduce new regulatory
considerations. However, it is important to also acknowledge that the utilization of GenAI in
medical devices should not fundamentally alter the core principles that have long governed
medical device regulation. The existing risk-based premarket and postmarket frameworks are
designed to support innovation and novel technologies in medical devices while ensuring device
safety and effectiveness. When technology introduces novel considerations, existing frameworks
and principles can be adapted in a targeted and risk-proportionate manner, rather than creating
separate technology-specific regulatory frameworks or taxonomies. This approach preserves
regulatory predictability, promotes global convergence, and ensures that oversight remains
centered on the clinical risks and benefits of the device rather than a specific technology.

FDA’s three-tier medical device classification framework and premarket programs, coupled with
established benefit-risk principles and postmarket responsibilities, remain appropriate and
applicable for assessing device risk, safety, and effectiveness throughout the lifecycle of GenAIenabled devices. These foundational aspects of FDA regulations ensure device classification and
risk assessment remain tethered to the device's intended use, whereas some concepts introduced
in the discussion paper, such as the proposed 2-axis framework, overlook this critical element of
device risk. Any adaptations to existing frameworks should remain technology-agnostic and be
grounded in intended use, device function, and clinical risk to avoid creating blanket frameworks
or expectations tailored to a particular technology.

Build upon existing foundation of risk assessment and device classification
Medical device manufacturers use robust established methods to evaluate risk across a wide
range of devices. Existing international consensus-based resources used for medical device risk
characterization, assessment, and management, such as ISO 14971 “Medical devicesApplication of risk management to medical devices” and the IMDRF N81 document titled
“Characterization Considerations for Medical Device Software and Software-Specific risk 2”,
offer a strong and more granular foundation for characterizing and managing risks associated
with GenAI-enabled device functions than the proposed 2-axis framework.

2 Available online: https://www.imdrf.org/documents/characterization-considerations-medical-device-software-and-softwarespecific-risk
FDA Dockets Management Docket No. FDA-2026-N-7874
October 19, 2026
Page 3 of 9

The proposed 2-axis framework is described as an “organizing heuristic” that “may affect FDA’s
scope of regulatory oversight, its evidentiary requirements in the premarket setting, and its
expectations surrounding postmarket monitoring and change control.” However, it is unclear
how the 2-axis model would be applied in practice as it appears to conflate regulatory device
classification, risk characterization, and risk management. The model omits several relevant
dimensions related to device classification and risk characterization, including intended use.
Additionally, the model is presented in the context of FDA-regulated GenAI device software
functions, but it also encompasses informational and non-directive activities that may not meet
the statutory definition of a device software function 3 and/or may fall within FDA's current
enforcement discretion policies 4. As such, the proposed 2-axis model should not be used as a
guide for classifying GenAI-enabled devices. Furthermore, the presence of GenAI or the degree
of autonomy should not be the sole drivers of assessing device risk. Rather, consistent with
existing practice, a device’s risks should be evaluated holistically with consideration to its
intended use and device-specific technological considerations to more fully contextualize clinical
use-related risks.

The 2-axis plot depicts a simple linear relationship where degree of device autonomy is the
primary driver of consequence. However, autonomy and directiveness, attributes presented in the
x-axis, may actually be considered risk modifiers that inform the overall risk assessment rather
than serve as standalone determinants of risk or consequence. In practice, the assessment of
device risk for an AI-enabled function depends on a range of factors (e.g., reversibility,
explainability, probability of harm, degree of model constraint) that are excluded from the plot
but would otherwise be considered when applying the principles outlined in ISO 14971 and the
N81document.

With regard to multi-turn conversational GenAI-enabled devices that the discussion paper
contends “…may migrate from providing “non-directive” information to “action-directing”
information…”, existing frameworks on device design, risk assessment, verification and
validation are adequate to ensure that any medical device, including GenAI-enabled medical
devices, continue to operate within the confines of its intended use and indications for use.
GenAI-enabled devices can be designed with appropriate, device-specific controls, such as
system prompts, guardrails, access controls, and other safeguards, to ensure outputs remain
within the device's intended use and defined conversational scope. Risk assessment and
performance evaluation should consider the effectiveness of these controls, as well as the
device's performance across representative and reasonably foreseeable conversational
trajectories.

3 Informational/non-directive outputs may not meet the statutory definition of a regulated medical device per section 520(o)(1)(E)

of the FD&C Act and the accompanying FDA guidance titled, “Clinical Decision Support Software.”
https://www.fda.gov/regulatory-information/search-fda-guidance-documents/clinical-decision-support-software
4 Informational outputs, including those intended to support a healthy lifestyle, may fall under FDA’s compliance policy for low

risk products intended to promote a healthy lifestyle. FDA’s policy is outlined in the guidance document titled, “General
Wellness: Policy for Low Risk Devices.” https://www.fda.gov/regulatory-information/search-fda-guidance-documents/generalwellness-policy-low-risk-devices
FDA Dockets Management Docket No. FDA-2026-N-7874
October 19, 2026
Page 4 of 9

Consistent with the Agency’s benefit/risk guidance documents for medical devices 5, it is
important that FDA’s discussions related to device risk assessment also include an assessment of
benefit. For example, autonomous systems may improve outcomes, particularly in environments
affected by specialist shortages or limited healthcare access. As such, device benefit is also an
important and relevant characteristic to consider in any discussion or model related to the
assessment of risk for GenAI-enabled devices.

CDRH has indicated 6 that a draft guidance titled “Risk Assessment and Evidentiary
Considerations for Generative Artificial Intelligence-Enabled Devices” is under construction and
that it intends to issue a draft guidance titled “Generative Artificial Intelligence-Enabled
Conversational Devices for Mental Disorders: Evidentiary Recommendations.” It is
recommended that FDA build upon the principles outlined in the aforementioned existing
resources and other established risk-characterizing principles in lieu of crafting new risk models
or unique regulatory concepts for GenAI-enabled medical devices. Future guidance documents
should include an explanation of FDA’s policies that are supplemented with clear definitions,
objective criteria, and illustrative examples to ensure consistent interpretation.

Premarket evaluation and evidentiary expectations should be calibrated to the device
We appreciate CDRH’s commitment to working collaboratively with stakeholders to identify
scientifically sound methods for evaluating GenAI-enabled devices that adhere to FDA’s least
burdensome principles 7. As with all medical devices, FDA’s premarket evidentiary expectations
for GenAI-enabled devices should be calibrated to the device’s intended use, technological
characteristics, and risks, ensuring performance evaluations are relevant and fit for purpose.
Existing medical device evaluation considerations including performance verification,
performance validation, cybersecurity, and usability should continue to play a central role. The
discussion paper contemplates a competency-based approach to premarket evaluation of GenAIenabled devices that includes device benchmarking and clinical confirmation. FDA’s objective
should not be to create new terminology or evidence frameworks for GenAI devices, but to
supplement existing types where the technology introduces novel considerations that are not
adequately characterized through established methods alone or when existing types would be too
burdensome if applied without adjustment.

5 FDA has issued several guidance documents that discuss considerations related to benefit/risk determinations for medical

devices, including but not limited to, “Consideration of Uncertainty in Making Benefit-Risk Determinations in Medical Device
Premarket Approvals, De Novo Classifications, and Humanitarian Device Exemptions,” “Factors to Consider When Making
Benefit-Risk Determinations in Medical Device Premarket Approval and De Novo Classifications,” and “Benefit-Risk Factors to
Consider When Determining Substantial Equivalence in Premarket Notifications (510(k)) with Different Technological
Characteristics.”
6 The guidance documents are included in CDRH’s list of proposed guidances for Fiscal Year 2027:

https://www.fda.gov/medical-devices/guidance-documents-medical-devices-and-radiation-emitting-products/cdrh-proposedguidances-fiscal-year-2027-fy-2027
7 FDA’s least burdensome principles related to medical devices are outlined in the guidance document titled, “The Least

Burdensome Provisions: Concept and Principles.” https://www.fda.gov/regulatory-information/search-fda-guidancedocuments/least-burdensome-provisions-concept-and-principles
FDA Dockets Management Docket No. FDA-2026-N-7874
October 19, 2026
Page 5 of 9

The inclusion of GenAI in a medical device should not automatically trigger a competency
assessment; the applicable benchmarking elements and clinical confirmation activities should be
selected according to the intended use, device capabilities, and their associated risks. Not all
benchmarking types or domains will be relevant to every device, and not all forms of clinical
confirmation described in the discussion paper will be necessary in every case. For example, in
some cases, GenAI device benchmarking can be conducted through least burdensome
approaches used for evaluating other types of AI-enabled medical devices, including statistical
performance evaluation and non-clinical testing. Similarly, for lower-risk GenAI-enabled
devices, established methods may provide sufficient evidence without additional clinical
confirmation activities tied to a competency-based framework.

The paper introduces the term "competency-based" evaluation as “…inspired…by how human
clinicians are evaluated and credentialed.” This description may inadvertently create confusion
by implying GenAI medical devices are inherently akin to human clinicians and are thus
evaluated in the same manner. Furthering the confusion, it appears that the “competency based”
approach is presented as a novel approach to medical device evaluation, yet some elements of
“benchmarking” and “clinical confirmation” described in the paper are related to existing
principles for medical devices, including analytical validation and clinical validation. FDA
should clarify that premarket evaluation of GenAI-enabled devices is not intended to replicate
the credentialing of clinicians, but rather to assess whether the device demonstrates it safely and
effectively performs within its intended use.

FDA should leverage and, when necessary, adapt premarket evidentiary expectations, methods,
and terminology rather than creating new parallel expectations and vocabularies specific to
GenAI. This will support a common understanding among stakeholders and avoid the
unnecessary burden of reconciling multiple overlapping taxonomies and terms.

Adapt PCCP implementation and other existing resources to improve utility and
applicability for GenAI-enabled devices
FDA’s Predetermined Change Control Plan (PCCP) authority is an appropriate mechanism for
managing postmarket changes to GenAI-enabled devices. Indeed, PCCP is a means to generate
regulatory efficiency for postmarket device changes for all types of devices. However, FDA’s
current implementation of this authority, as outlined in guidance 8, may be too restrictive to
achieve its full potential. For GenAI-enabled devices, in particular, where future modifications
may not be fully anticipated in advance, PCCP implementation would benefit from greater
flexibility regarding the scope of permissible changes and shifting FDA’s emphasis from the
prespecified change to a more agile approach. For example, instead of requiring manufacturers to
pre-specify each individual change in the Description of Modifications section of the PCCP,
FDA could allow them to define broader boundaries and guardrails that describe a range of

8 FDA’s final guidance titled “Marketing Submission Recommendations for a Predetermined Changed Control Plan for Artificial

Intelligence-Enabled Device Software Functions” and the draft guidance titled “Predetermined Change Control Plans for Medical
Devices” both apply limitations and constraints on the usage and implementation of PCCP authority that appear inconsistent with
statute and restrict the practical utility of PCCPs for all medical devices.
FDA Dockets Management Docket No. FDA-2026-N-7874
October 19, 2026
Page 6 of 9

allowable changes. FDA can preserve confidence in its oversight by ensuring the PCCP contains
appropriate risk controls, evaluation and validation methodology, and performance thresholds in
the Modification Protocol and Impact Assessment portions of the PCCP. This approach would
retain appropriate FDA regulatory oversight and manufacturer accountability while avoiding an
expectation to foresee the exact technical form of every future modification, which is inherently
impractical for most GenAI-enabled systems. These suggested modifications to PCCP
implementation are offered in the context of the GenAI discussion paper, however these
improvements can be applied to PCCP implementation more broadly to support regulatory
efficiency for all devices.

We encourage FDA to consider the recommendations regarding PCCP implementation outlined
in the UK’s recently published paper titled, “National Commission into the Regulation of AI in
Healthcare-Recommendations for a future regulatory framework 9” which describes creating an
approach “… to PCCPs that allows for manufacturers to establish boundaries to define the scope
of allowable change (rather than only allowing specific modifications to be prespecified),
including clear approaches for adaptive AI-enabled medical devices, while maintaining a
device’s intended use.” As a general matter, FDA should continue to evaluate and evolve its
implementation of the PCCP authority through regular collaboration with stakeholders. Such
collaboration could support timely updates to guidance and development of other types of
regulatory support tools or guides to ensure PCCP is used effectively for GenAI and other
rapidly evolving and emerging technologies.

In addition to implementing the PCCP authority with greater flexibility, FDA should explore
additional opportunities to support agile regulation of novel technologies. For example, FDA
should consider updating the “Deciding When to Submit a 510(k) for a Software Change to an
Existing Device 10” guidance to offer more clarity and risk-based flexibility on the types of
changes that do not necessitate a new 510(k) clearance for GenAI-enabled devices and other
software modifications.

A combination of approaches, including a more practical implementation of PCCPs, updating
existing guidance documents, collaborating with stakeholders, and harmonizing with best
practices used by other global regulators can generate the necessary regulatory efficiency,
predictability, and flexibility necessary to promote innovation without adversely impacting
patient safety.

9 The National Commission paper is available online: https://www.gov.uk/government/publications/national-commission-intothe-regulation-of-ai-in-healthcare-recommendations-for-a-future-regulatory-framework. The MHRA’s response to the paper,
which indicates intention to implement all recommendations in the paper is also available online:
https://www.gov.uk/government/publications/government-response-to-the-national-commissions-recommendations-on-theregulation-of-ai-in-healthcare
10 Available online: https://www.fda.gov/regulatory-information/search-fda-guidance-documents/deciding-when-submit-510ksoftware-change-existing-device
FDA Dockets Management Docket No. FDA-2026-N-7874
October 19, 2026
Page 7 of 9

Postmarket surveillance is a shared responsibility with distinct responsibilities for
manufacturers based on device-specific considerations
The discussion paper considers rebalancing premarket evidentiary expectations and postmarket
monitoring as a potential means to support timely patient access while maintaining appropriate
safeguards. Greater reliance on postmarket evidence may be appropriate for certain GenAIenabled medical devices under well-defined conditions. FDA already has established
mechanisms for applying additional postmarket oversight where appropriate, including special
controls for De Novo devices and postmarket study or reporting requirements for PMA devices.
We generally support FDA continuing to utilize these existing authorities, where warranted, to
facilitate patient access to devices that have demonstrated safety and effectiveness while
allowing remaining uncertainties to be addressed through risk-proportionate postmarket controls.
The appropriate balance should be established during premarket review, including the expected
evidence, monitoring obligations, and assessment criteria. Postmarket evidence should not be
used as a substitute for FDA’s premarket determination of reasonable assurance of safety and
effectiveness for the device.
FDA should avoid prescribing a uniform postmarket monitoring approach across all GenAIenabled devices. Consistent with the current total product lifecycle paradigm for all medical
devices, manufacturers implement risk-based monitoring programs that are tailored to the
device's intended use, clinical significance, potential consequences of incorrect outputs, and the
likelihood that performance degradation could adversely affect safety or effectiveness.
Monitoring plans, including performance degradation monitoring and/or periodic rebenchmarking for GenAI-enabled devices may be established, if appropriate, during the initial
authorization process, allowing flexibility for manufacturers to select the most appropriate
surveillance methods for their specific device and use case. The appropriate monitoring
activities, cadence, and reassessment triggers should be determined based on device-specific
factors such as intended use and risk profile. Postmarket monitoring approaches should also
account for applicable privacy, security, and data-governance considerations, particularly where
monitoring involves collection or retention of user interactions or device outputs.

Clinicians, healthcare institutions, professional societies, standards-setting bodies, patients, and
other stakeholders can provide valuable real-world information and feedback to support
postmarket monitoring. However, manufacturers must retain responsibility for postmarket
surveillance, including establishing mechanisms to collect and analyze performance data,
evaluating safety signals, conducting investigations, implementing corrective and preventive
actions when necessary, and maintaining the ongoing safety, effectiveness, and lifecycle
management of the device. Under existing quality system regulations, manufacturers are also
responsible for controls that are inherent to the device, model governance, performance
monitoring capabilities, labeling, risk mitigations, and software updates. Clinicians and
healthcare institutions may contribute, as appropriate, through sample‑based review, adverse
event reporting, or real‑world feedback. Professional societies and standards bodies may help
define, for example, best practices, evaluation methodologies, and reporting formats. Stakeholder
participation should be structured to support, not replace, manufacturer-led oversight to maintain
FDA Dockets Management Docket No. FDA-2026-N-7874
October 19, 2026
Page 8 of 9

clear accountability. Expectations should also be proportionate to intended use and clinical risk,
rather than imposing uniform monitoring requirements across all GenAI-enabled devices.

FDA should also recognize that existing quality-system, risk-management, complaint-handling,
and vigilance processes provide effective mechanisms for identifying many performance
degradation signals and initiating corrective actions when needed. In some cases, periodic rebenchmarking or focused performance reviews may be appropriate, whereas in other cases,
routine surveillance and event-driven reassessment may provide greater value. Overall, FDA's
framework should remain flexible, allowing manufacturers to employ a combination of
monitoring approaches proportionate to the specific risks presented by the device while ensuring
continued safety and effectiveness throughout its lifecycle.

FDA should retain sole oversight of regulated medical devices
The discussion paper contemplates the potential role and utility of independent third-party
involvement in the context of applying a competency-based approach to device evaluation.
FDA’s existing oversight of AI-enabled medical devices renders third-party oversight of device
performance redundant and impractical. If third-party device evaluation were linked to the
regulatory process, their use would effectively become mandatory, increasing burdens for
manufacturers who already evaluate their devices in accordance with best practices like those
cataloged in FDA-recognized international consensus standards.

We are also concerned that third-party evaluators, such as assurance laboratories, may utilize
proprietary methods and metrics that would lack transparency to developers, patients, and
providers. There are also significant concerns about their handling of sensitive training and
testing data. Mandating manufacturers to share proprietary information about AI-enabled devices
with external laboratories raises confidentiality, intellectual property (IP), and security issues as
variability in security practices for third-party laboratories could expose or jeopardize sensitive
manufacturer or patient information. Implementing a third-party framework of laboratories or
evaluators creates, at best, redundancy with existing FDA oversight while potentially increasing
healthcare costs, slowing innovation, and limiting access to valuable medical devices, without
offering clear benefit to clinicians or patients.

Rather than create a new framework of third-party GenAI device evaluators, we encourage FDA
to participate in the development and timely recognition of accredited and consensus-based
standards for safety and quality assurance processes. The longstanding practice of relying on
collaboratively developed international consensus standards promotes a deeper understanding of
both the benefits and risks of these technologies while improving transparency. Enabling
manufacturers to utilize consensus standards allows for robust quality and safety assurance that
maintains data security and IP confidentiality. This approach supports patient safety and
innovation by minimizing the need for external assessments, which could introduce added costs,
regulatory delays, and potential security vulnerabilities.
FDA Dockets Management Docket No. FDA-2026-N-7874
October 19, 2026
Page 9 of 9

We appreciate the opportunity to engage in this important discussion and welcome the
opportunity for continued collaboration. Please feel free to contact us with any questions.

Sincerely,

Geeta Pamidimukkala, MS Zack Hornberger
GPamidimukkala@advamed.org zhornberger@advamed.org
Vice President, Technology & Regulatory Associate Vice President, Digital Health &
Affairs Imaging Technology
AdvaMed AdvaMed Medical Imaging Division