ACRO (Karen Noonan)

October 19, 2026

Dr. Michelle Tarver, Director, Center for Devices and Radiological Health
Food and Drug Administration
10903 New Hampshire Avenue
Silver Spring, MD 20993

RE:ACRO comment on FDA CDRH Discussion Paper Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion Paper and Request for Feedback
[Docket FDA-2026-N-7874]

Dear Dr. Tarver,

ACRO is pleased to submit this comment on the CDRH Discussion Paper.
Our feedback is attached here as a PDF. Please see attached.

Respectfully submitted,
Karen Noonan
Senior Vice President, Global Regulatory Policy, ACRO

ATTACHMENT: ACRO Comment (PDF)

October 19, 2026

Dr. Michelle Tarver, Director, Center for Devices and Radiological Health
Food and Drug Administration
10903 New Hampshire Avenue
Silver Spring, MD 20993

RE: ACRO comment on FDA CDRH Discussion Paper Considerations for the Regulation of Generative AIEnabled Medical Devices: Discussion Paper and Request for Feedback
[Docket FDA-2026-N-7874]

Dear Dr. Tarver,

Founded in 2001, the Association of Clinical Research Organizations (ACRO) is a non-profit trade association
representing the world’s leading clinical research and technology organizations, which provide specialized
services that are integral to the development of drugs, biologics, and medical devices that enable patients to
live longer, healthier, and more productive lives. ACRO members provide a wide range of specialized services
across the entire spectrum of drug development—from preclinical, proof of concept, and first in human
studies through post-approval, pharmacovigilance, and health data research. ACRO member companies
employ more than 470,000 people worldwide and conduct research in every global region.

ACRO is pleased to provide feedback on this CDRH Discussion Paper and commends CDRH for seeking early
stakeholder input. Because generative artificial intelligence (“GenAI”) is evolving rapidly, ACRO encourages
CDRH to favor flexible, pragmatic, principles-based approaches that provide sponsors with clarity and
predictability while avoiding recommendations that are impractical, unduly burdensome, or likely to become
quickly outdated. Our key messages are set out below.

1. ACRO welcomes the two-axis risk framework, which aligns with FDA’s approach for drugs and biologics
CDRH proposes a two-axis risk framework that considers the device’s “activity” (ranging from
informational/non-directive to fully autonomous action-taking) and the “consequences” of relying on an
incorrect output (ranging from limited to severe). This CDRH two-axis risk framework builds upon, and
converges with, the two-axis framework outlined in the 2025 FDA draft guidance on Considerations for the
Use of Artificial Intelligence to Support Regulatory Decision-Making for Drug and Biological Products, which
presents a two-axis Model Risk Matrix based on “model influence” and “decision consequence.” ACRO
believes this alignment makes sense. The underlying principles (how much an AI output drives a decision or
action, and the consequences if that output is wrong) apply irrespective of whether the AI is part of a medical
device or supports drug and biologic development. A consistent approach across FDA centers will also benefit
sponsors, CROs, and technology developers that work across product types.

ACRO encourages CDRH to maintain the framework as a flexible, organizing heuristic within existing device
classification, benefit-risk, and risk-management principles, rather than as a new GenAI-specific classification
scheme. Other factors, such as the detectability and reversibility of errors, human oversight, and downstream
safeguards, may also be relevant and could be treated as contextual considerations rather than additional
fixed criteria. Likewise, the intended user (patient, generalist, or specialist) may be relevant context but
should not, on its own, determine a function’s risk. ACRO would welcome additional discussion and
communication from FDA regarding how CDRH, CDER, and CBER will keep their frameworks aligned as each is
finalized.

601 New Jersey Ave NW, Suite 350, Washington DC 20001 | info@acrohealth.org | www.acrohealth.org
2. ACRO supports a technology-neutral approach that builds on existing FDA frameworks
ACRO strongly supports CDRH’s statement that FDA regulates medical devices, not GenAI as such. The use of
GenAI should not, by itself, determine a device’s classification, regulatory pathway, or evidentiary
requirements; these should continue to rest on intended use, clinical context, and risk. Future policy should
build on existing frameworks, including those for AI-enabled device software functions, PCCPs, quality
systems, cybersecurity, and Good Machine Learning Practice, with targeted additions only where GenAI
introduces risks those frameworks do not already address. The same principle should apply to agentic
functionality, which should be assessed based on what the system is permitted to do and the consequences
of error rather than the “agentic” label, and to multi-function products, where oversight should continue to
apply at the device-function level.

3. ACRO supports flexible and risk-proportionate premarket evaluation
ACRO welcomes CDRH’s recognition that conventional testing approaches may not suit GenAI-enabled
devices and that clinical confirmation may not require a prospective clinical study in every case. The
competency-based approach of benchmarking followed by clinical confirmation may be useful for some
devices. However, given how quickly GenAI is evolving, ACRO encourages CDRH to present it as one of several
acceptable approaches rather than a required sequence or checklist. In particular, ACRO recommends that
CDRH clarify in future communications that (1) benchmarking elements are a menu from which sponsors
select based on intended use and risk; (2) sponsor-developed benchmarks and datasets are acceptable when
shown to be fit for purpose; (3) independent third-party testing or adjudication is an option rather than a
default requirement; and (4) the choice of comparator (e.g., a clinical reference standard, standard of care,
unaided user, or human-device team performance) follows the intended use rather than defaulting to
physician performance.

ACRO members have extensive experience designing and conducting clinical studies for devices, drugs, and
biologics, and stand ready to work with CDRH on practical, scientifically sound approaches to clinical
confirmation, including the appropriate use of real-world data and data collected outside the United States.

4. ACRO supports risk-based, pragmatic postmarket monitoring and change control
ACRO supports a total product life cycle approach and agrees that postmarket evidence can complement, but
should not replace, the premarket evidence needed to provide a reasonable assurance of safety and
effectiveness. Re-benchmarking, sample-based clinician review, and drift monitoring may each be useful, but
none should be universally required; manufacturers should be able to set the scope, frequency, and triggers
for monitoring based on device risk and identified failure modes. Similarly, evaluation after a modification
should be proportionate to its reasonably foreseeable impact, and PCCPs should be able to define a bounded
range of anticipated changes without specifying every future technical implementation. ACRO requests FDA
to clarify how it will distinguish expected variability in generative outputs, on the one hand, from
performance degradation that affects safety or effectiveness, on the other. We ask the Agency to clarify its
expectations for manufacturers using third-party foundation models, which should reflect the information
and controls reasonably available to them.

5. ACRO supports voluntary Foundation Model Master Files
ACRO sees value in voluntary Foundation Model Master Files as a way to reduce duplicative submissions and
give FDA reviewers consistent information. Participation should remain voluntary; the absence of a Master
File should not disadvantage a sponsor that can support its submission with device-level evidence; and a
Master File should not be viewed as FDA qualification or endorsement of the underlying model.

601 New Jersey Ave NW, Suite 350, Washington DC 20001 | info@acrohealth.org | www.acrohealth.org
ACRO also encourages CDRH to continue working with international regulators, including through the
International Medical Device Regulators Forum (IMDRF), to promote convergence on terminology, validation,
and change-management expectations. Given the pace of change in GenAI, ACRO believes that high-level,
principles-based policy, supplemented by illustrative examples and early engagement through the QSubmission program, will serve patients better than detailed, prescriptive requirements that may quickly
become outdated.

ACRO appreciates this opportunity to comment and welcomes continued engagement with CDRH on this
important topic. Please do not hesitate to contact ACRO (knoonan@acrohealth.org) if we can answer any
questions or provide additional information.

Respectfully submitted,

Karen Noonan
Senior Vice President, Global Regulatory Policy

601 New Jersey Ave NW, Suite 350, Washington DC 20001 | info@acrohealth.org | www.acrohealth.org