Personalized Medicine Coalition (Cynthia A. Bens)

See attached.

Gina Ross Murdoch, M.B.A
President and CEO October 19, 2026
Lauren Silvis, J.D.(Chair)
Tempus Dockets Management Staff
Sarah Hersey, M.S., M.B.A., R.A.C. U.S. Food and Drug Administration
(Vice Chair)
Bristol Myers Squibb 5630 Fishers Lane
Richard Knight, M.B.A. (Treasurer)
Room 1061
American Association of Kidney Rockville, MD 20852
Patients

Brian Caveney, M.D. (Secretary)
Labcorp RE: Considerations for the Regulation of Generative AI-Enabled Medical
Lincoln Nadauld, M.D., Ph.D. (Past
Devices Discussion Paper [FDA-2026-N-7874]
Chair)
Valinor
Dear Sir/Madam:
Okan Ekinci, M.D., M.B.A.
Roche Information Solutions
Roche Diagnostics The Personalized Medicine Coalition (PMC), a multi-stakeholder group comprising
Helmy Eltoukhy, Ph.D. more than 150 institutions from across the health care spectrum, thanks the U.S.
Guardant Health
Food and Drug Administration (FDA) for the opportunity to submit comments
Andrea Ferris, M.B.A.
LUNGevity
related to the discussion paper titled Considerations for the Regulation of
Generative AI (GenAI)-Enabled Medical Devices. i PMC believes that emerging and
Megan P. Hall, Ph.D.
GRAIL Inc. potential uses of AI may begin to improve patient experiences with health care and
Ryan Hutto, M.A.
help address challenges to the clinical adoption of personalized medicine. ii PMC
Pfizer commends the FDA for recognizing regulatory challenges for GenAI-enabled
James Lillard, Ph.D., M.B.A. medical devices and contemplating ways to advance regulatory approaches for these
Morehouse School of Medicine
devices. We appreciate FDA’s early engagement with stakeholders on possible
Howard L. McLeod, Pharm.D.
Clarified Precision Medicine
approaches to risk assessment, premarket evaluation, postmarket monitoring, and
other topics relevant to the regulation of GenAI-enabled devices. PMC shares
Sharon Moulis, Ph.D.
Independent FDA’s ultimate goal of enabling a nimble regulatory approach that allows patients
Elizabeth O’Day, Ph.D.
timely access to safe and effective GenAI-enabled medical devices.
Olaris, Inc.

Omar Perez, Ph.D., R.A.C. Personalized medicine is a rapidly evolving field in which physicians use diagnostic
Independent
tests and individual details about a person and their health to determine which
Kristin Ciriello Pothier
KPMG
medical treatments will work best for each patient or use medical interventions to
alter molecular mechanisms that impact health. By combining data from diagnostic
Prasanth Reddy, M.D.
Mercy BioAnalytics tests with an individual’s medical history, circumstances, and values, health care
Cecelia Schott, Pharm.D., M.B.A.
providers can develop targeted treatment and prevention plans with their patients.
GSK

Anthony Sireci, M.D., M.Sc. Personalized medicine is helping to shift patient and provider experiences away
Eli Lilly and Company
from trial-and-error toward a more streamlined process for making clinical
Andrea Stevens, Ph.D.
Johnson & Johnson
decisions, which leads to improved patient outcomes, a reduction in unnecessary
treatment costs, and better patient and provider satisfaction. PMC and its members
Apostolia-Maria Tsimberidou, M.D.,
Ph.D. are leading the way in personalized medicine and in developing evidence showing
M.D. Anderson Cancer Center,
University of Texas
how patients and the health care system can benefit from appropriate testing and
tailored treatment as soon as possible during their clinical experiences.
Michael Ybarra, M.D.
PhRMA
Statement of Neutrality

PMC’s members may present their own responses to the FDA’s discussion paper titled Considerations for the
Regulation of Generative AI-Enabled Medical Devices and actively advocate for those positions. PMC’s
response is designed to provide feedback so that the general concept of personalized medicine can advance, and
is not intended to impact adversely the ability of individual PMC members, alone or in combination, to submit
separate responses to FDA on this discussion paper.

PMC advocates for public policies that can help AI realize its full potential for improving patient access to
personalized medicine and for making the health care system more efficient across disease areas. Many of the
characteristics that may make GenAI useful – including flexibility, natural-language interaction,
personalization, and multimodality – can also introduce new challenges for evaluating performance and
identifying failures. PMC generally supports a regulatory framework that is appropriately risk-based,
scientifically rigorous, predictable, and sufficiently flexible to accommodate technological change. In
developing a risk-based framework for oversight, regulators should consider differing levels of explainability in
how AI models arrive at their outputs. Regulators should also allow adaptive regulatory approaches that allow
for iterative development and continuous improvement of AI-based tools and technologies.

The following comments were informed by PMC’s Policy Principles on AI, developed in collaboration with
PMC’s AI Working Group. iii We encourage the FDA to consider these recommendations as it further develops
its approach to GenAI and when developing any future guidance documents relevant to AI-enabled medical
products that support personalized medicine.

Considerations for Risk Assessment

There are many use cases for AI in health care relevant to personalized medicine and, depending on the use
case, developers may be subject to different regulatory requirements. PMC’s focus is on ensuring the right
framework for AI is used throughout the total product lifecycle. GenAI is designed to create new content by
leveraging a large corpus of data used for its training. These models learn the underlying distribution of the
input data and can produce novel outputs such as images, text, or audio. Common applications include image
generation, text generation, and music or speech synthesis. In health care applications, this includes generating
synthetic medical data, creating personalized treatment plans, and simulating patient responses to different
health care therapies. iv The right framework for regulatory oversight of AI will ensure confidence in the
performance and use of AI and mitigate risks to patients while continuing to foster innovation.

In this context, the two-axis risk assessment framework in Section IV of FDA’s discussion paper, in particular
the recognition that a function's level of independent activity, moderated by the severity of the consequences of
an erroneous result, should inform regulatory expectations. These dimensions begin to distinguish between tools
that provide information, tools that direct users toward a particular action, and tools that take action with limited
or no human intervention. They also begin to recognize that similar levels of autonomy may carry different risks
depending on the clinical consequences. However, we urge the FDA to reevaluate the factors it suggests should
alter a function’s risk classification. Specifically, FDA claims that GenAI-enabled in vitro diagnostic
measurement and signal processing functions raise increased concern as their outputs typically cannot be
independently assessed by the user. Because such functions do not direct users to perform an action, FDA
should not automatically propose to treat them as higher risk. Regulatory oversight should take the form of an
appropriate risk-based and flexible approach. FDA’s framework would benefit from considering whether the
intended user can understand the basis and limitations of an output, recognize uncertainty, and obtain meaningful
review before acting.

Regulatory Treatment Based on Intended User
The discussion paper raises questions about whether GenAI-enabled patient-facing functions may present
different risk considerations from GenAI-enabled health care provider (HCP)-facing functions. FDA also asks a
related question with respect to GenAI-enabled functions intended for use by HCPs without specialty-level
training in the clinical domain for which the function is intended. FDA’s framework would be best applied
based on the intended use of the function itself, rather than by assigning a level of risk based on who deploys it.

GenAI can generate highly personalized and persuasive outputs that patients may reasonably interpret as
tailored medical advice. FDA could consider directiveness along a continuum that accounts for specificity,
personalization, confidence, context, repetition across a conversation, and whether the model adapts its
recommendation after receiving additional patient information. A generic explanation of treatment options, for
example, presents different considerations than a response that integrates an individual’s symptoms and
medications and recommends a specific course of action. Therefore, the relevant regulatory question is whether
the intended user has an opportunity to understand the output’s limitations, recognize uncertainty or error, and
obtain appropriate review before a consequential action occurs. Safeguards could be put in place depending on
an intended use. These may include uncertainty communication, source traceability, explicit boundaries on
intended use, escalation pathways, confirmation steps, and human review for higher-consequence
recommendations.

A similar approach could be considered to address differences between HCPs with and without specialty
training in a specific domain. A GenAI-enabled function that responsibly extends access to specialty knowledge
may offer meaningful patient benefit, particularly in settings where specialist access is limited. Professional
credentials alone should not determine risk. Risk may increase when safe use depends on knowledge the
intended user cannot reasonably be expected to possess, when an output falls outside the user’s usual scope of
practice, or when specialist-level conclusions are presented without sufficient information about uncertainty,
limitations, or the need for consultation. FDA should consider the knowledge reasonably available to the
intended user and whether the GenAI-enabled device provides adequate context and pathways for referral or
consultation.

Competency-Based Approach

PMC understands that the FDA is considering a competency-based approach to premarket evaluation of GenAIenabled devices. Exhaustive testing of every possible input and output is unlikely to be practical for open-ended
GenAI systems. A structured set of competencies may provide a scalable way to evaluate safety behavior,
clinical proficiency, communication, robustness, subgroup performance, and agentic capabilities.

As described in Section V of the discussion paper, the competency-based approach would consist of nonclinical benchmarking and clinical confirmation. Under such an approach, the final user-facing device, as
configured and intended to be deployed for real-world use – and not the foundation model standing alone or
other isolated subcomponent – would be evaluated. Benchmarking can efficiently test broad capabilities and
known failure modes, while clinically representative evidence can help establish whether those capabilities
translate into safe and effective real-world performance. This approach is worth exploring further for its ability
to be tailored to a device’s intended use and be proportionate to its risk.

Postmarket Monitoring
To assess potential benefits and risks to patients comprehensively, FDA’s evolving framework should consider a
GenAI-enabled device’s total product lifecycle. This approach could include the use of real-world monitoring to
identify and address issues or challenges that may emerge with implementation of AI, as well as identify
realized benefits that could expand a product’s utility and permissible uses.

Variable outputs, changing deployment environments, model updates, and dependencies on third-party
foundation models make ongoing evaluation of GenAI-enabled devices particularly important. The FDA’s total
product lifecycle approach, as described in Section VI of the discussion paper, is a critical starting point for
managing modifications, real-world performance monitoring, and continuous risk assessments. Within this
framework, Predetermined Change Control Plans can enable the controlled evolution of AI models in the
postmarket setting.

In PMC’s response to FDA’s draft guidance on AI-enabled device software functions we previously encouraged
robust postmarket monitoring and verification of AI-enabled devices. We recommend a similar approach in the
context of this framework. Emphasizing ongoing quality systems, postmarket surveillance, user feedback, and
adverse event monitoring as components of FDA’s lifecycle framework is consistent with other stakeholder
work in this area. v Postmarket plans for GenAI-enabled devices may include relevant elements such as
performance thresholds, subgroup analyses, sampling methods, reassessment cadence, triggering events,
escalation procedures, and the actions expected when thresholds are crossed. Triggering events may include
clinically meaningful changes to the foundation model, prompts, retrieval sources, guardrails, orchestration
logic, user interface, intended population, deployment setting, or connected tools.

FDA has emphasized that it is important for developers to consider how to effectively evaluate and monitor
GenAI-enabled devices once they are deployed. The discussion paper states that FDA “is considering whether it
is appropriate to accept greater premarket uncertainty regarding a GenAI-enabled device’s benefit-risk profile
through greater reliance on postmarket monitoring.” Such an approach may be appropriate for some functions
where errors are readily detectable and reversible, monitoring can identify deterioration promptly, and there is a
reliable mechanism to intervene before harm accumulates. Where FDA accepts greater premarket uncertainty,
there should be clear data-quality expectations, preservation of relevant logs and version information, and a
defined plan for how postmarket evidence will address the remaining uncertainty.

Conclusion

PMC appreciates FDA’s interest in advancing timely patient access to safe and effective Gen AI-enabled
medical devices. We look forward to working with you and your colleagues to further shape concepts in the
discussion paper so that the current and future role of AI in advancing personalized medicine is considered in
any further action taken by the FDA. If you have any questions about the content of this letter, please contact
me at cbens@personalizedmedicinecoalition.org or 202-499-0986.

Sincerely,

Cynthia A. Bens
Senior Vice President, Public Policy
i U.S. Food and Drug Administration. Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion
Paper and Request for Feedback. August 2026. https://www.fda.gov/media/194242/download. (accessed October 14, 2026).
ii
C.S. Ajmal, S. Yerram, V. Abishek, et al. "Innovative Approaches in Regulatory Affairs: Leveraging Artificial Intelligence and
Machine Learning for Efficient Compliance and Decision-Making." AAPS Journal 27 (2025): 22. https://doi.org/10.1208/s12248-02401006-5 (accessed October 14, 2026).
iii
Personalized Medicine Coalition. Response to White House Office of Technology Policy Request for Information on the
Development of an AI Action Plan. March 14, 2025. https://www.personalizedmedicinecoalition.org/wpcontent/uploads/2024/11/PMC-Comment-Letter-OSTP-RFI-AI-Action-Plan-FINAL.pdf. (accessed October 14, 2026).
iv
Advanced Medical Technology Association, The Role of Artificial Intelligence (AI) in Healthcare, November 11, 2024,
https://www.advamed.org/member-center/resource-library/the-role-of-artificial-intelligence-ai-in-healthcare/. (accessed October 14,
2026).
v
Advanced Medical Technology Association. Response to Congressional RFI on AI in Health Care. May 9, 2024,
https://www.advamed.org/industry-updates/news/advamed-responds-to-congressional-rfi-on-ai-in-health-care/ (accessed October 14,
2026).