Oura (Ricky Bloomfield, MD, CMO)

Please see the attached formal comment letter submitted by Oura regarding Docket No. FDA-2026-N-7874.

415 Kearny Street,
San Francisco, California 94108
ouraring.com

September 8, 2026

Submitted electronically via Regulations.gov

Re: Docket No. FDA-2026-N-7874; Considerations for the Regulation of Generative
AI-Enabled Medical Devices: Discussion Paper and Request for Feedback

Dear Dr. Tarver and Dr. Abramson,

Oura appreciates the opportunity to comment on the Digital Health Center of Excellence's
discussion paper on the regulation of generative AI (GenAI)-enabled medical devices, and we
commend the Center for engaging stakeholders early and transparently as it develops its
approach.
​
Founded in 2013, Oura is a health technology company and maker of the Oura Ring, a smart
ring that continuously tracks 50+ health metrics to deliver personalized health data, insights, and
daily guidance to more than 5 million members worldwide. Oura sits at the intersection of the
issues the paper raises: our members generate rich longitudinal health data through their rings,
and they increasingly expect AI-enabled features that help them understand what that data
means.

Our comments are deliberately not exhaustive. Rather than responding to each of the paper's
questions, we focus on the aspects most relevant to wearables: the proposed considerations for
risk assessment and (1) what this framework, if adopted, would mean for the regulatory status of
informational, non-directive GenAI-enabled functions offered directly to consumers, (2) how this
framework would interact with the Agency's existing wellness policy, and (3) how the Agency
should ensure that regulatory treatment of GenAI-enabled functions depends on the intended
use of the function, not on who deploys it.

1. Informational, non-directive functions should be treated as non-device functions or
placed under enforcement discretion.

We support the two-axis risk assessment framework in Part II of the discussion paper, and in
particular the recognition that a function's level of independent activity, moderated by the
severity of the consequences of an erroneous result, should drive regulatory expectations.
Under this proposed rubric, FDA acknowledges that informational, non-directive functions with
limited consequences to users in the event of an incorrect output pose the lowest level of risk.
We encourage the Agency to clarify in guidance or through rulemaking that FDA
considers GenAI-enabled functions that provide informational, non-directive output to
users to be non-devices or, at minimum, to place such functions under enforcement
discretion because these functions present a low-risk to users when the output does not
work as intended. This approach would be consistent with the Agency's longstanding
treatment of similar low-risk digital health products such as symptom checkers and
functions intended to help users self-manage their diseases or conditions without
providing specific treatment or treatment suggestions.1

Informational, Non-directive GenAI-enabled Functions are Low Risk

A function that helps a user understand their own data, provides general context about the
meaning of a wearable reading and how it may relate to the user’s medical history and
self-described symptoms, and may encourage the user to consult a healthcare professional
exemplifies the lowest-level risk output of GenAI-enabled functions. The information provided is
purely educational, does not direct a specific treatment, and informs the user that a healthcare
professional can provide more information. The provision of educational information presents
limited risk should the information be incorrect because no treatment is directed and the user is
encouraged to consult a healthcare professional for evaluation or treatment.

Low Risk Devices Have Historically Been Under Enforcement Discretion

Given the low-risk nature of informational, non-directive functions and the limited consequences
an incorrect output would pose to a user, FDA should expressly place such functions under
enforcement discretion. This would be consistent with FDA’s approach articulated in the Policy
for Device Software Functions and Mobile Medical Applications guidance; i.e., that FDA will
focus its regulatory oversight on those devices that pose a risk to user safety when they do not
operate as intended. FDA has for decades focused its oversight on software functions that
transform a general-purpose computing or mobile platform into a regulated medical device that
is used in or provides clinical treatment. By contrast, FDA has declined to actively regulate
symptom checkers and self-management software tools. GenAI-enabled software that provides
similar functionality, helping users understand their symptoms and self-manage diseases or
conditions should be regulated similarly. The fact that the output is produced by a generative
model rather than a decision tree does not change the function performed or the risk to the user,
and therefore, the regulatory outcome should not change either.

The principle that informational recommendations should not be within FDA’s purview is also
consistent with the congressional intent behind the 21st Century Cures Act which removed from
FDA’s jurisdiction, among other things, clinical decision support software that provides
informational recommendations to a healthcare professional. This is the case even when the
output addresses a specific disease or condition, so long as the software function is intended to
See U.S. Food & Drug Admin., Ctr. for Devices & Radiological Health, Policy for Device Software
Functions and Mobile Medical Applications: Guidance for Industry and Food and Drug Administration
Staff (Sept. 28, 2022), https://www.fda.gov/media/80958/download.
support decision making while allowing the healthcare professional to independently review the
basis for those recommendations. FDA’s existing software guidance documents are clear that
the dispositive factor is not the subject matter of the output but its non-directive character and
the distance between the output and the treatment provided-- i.e., the software informs a
decision rather than making or directing one.

2. The Agency should clarify how the framework intersects with the General Wellness
Policy.

Oura encourages the Agency to clarify how the framework contemplated in this discussion
paper interacts with its General Wellness: Policy for Low Risk Devices guidance.2 Without
additional clarity, the two documents could be read to pull in different directions. The discussion
paper appears to contemplate GenAI-enabled functions that reference specific diseases and
conditions as occupying the low-risk end of the framework so long as the output is informational
and non-directive, such that FDA might consider exercising enforcement discretion. The
wellness guidance, by contrast, limits the informational output that such products can provide
while remaining within the scope of that policy’s enforcement discretion. Specifically, general
wellness products cannot reference particular diseases or conditions except through claims that
relate to reducing the risk of or living well with certain chronic diseases or conditions, nor can a
general wellness product provide guidance intended to inform medical management. Additional
clarity is needed around where wellness ends and device regulation begins for an AI feature
that helps a user interpret their own wearable data.

We recommend the Agency use this framework to establish a brighter line: an
informational, non-directive reference to a disease or condition made by a GenAI-enabled
function solely to provide context on the user's own wearable data remains a low-risk
function and does not, by itself, convert a wellness product into a medical device. If the
Agency nonetheless concludes that such a function meets the device definition, it should
exercise enforcement discretion, provided the output remains non-directive and
encourages the user to seek a confirmatory test or consult a healthcare professional. As
discussed below, that bright line should apply equally to non-GenAI informational,
non-directive functions.

Relatedly, the Agency should confirm that a prompt to consult a healthcare professional does
not itself take a function outside the wellness policy. The wellness guidance makes clear that
even if an output falls outside the ranges appropriate for general wellness use, the product may
still be considered a wellness product if it notifies users that evaluation by a healthcare
professional may be helpful. An informational action-directing GenAI-enabled function with that
feature should be treated consistently with the wellness guidance and not be considered a
device on that basis.
U.S. Food & Drug Admin., Ctr. for Devices & Radiological Health, General Wellness: Policy for Low Risk
Devices: Guidance for Industry and Food and Drug Administration Staff (Jan. 6, 2026),
https://www.fda.gov/media/90652/download.
3. The Agency should ensure that Regulatory treatment depends on the function’s
intended use, not on who deploys it. LLMs that interpret health and wellness data,
including wearable data, and provide interpretive guidance to users should be held to the
same standard by the Agency, regardless of who deploys them.

Consider the practical scenario we have in mind: a member whose wearable data shows
persistently elevated overnight heart rate and disturbed sleep asks the company's AI feature
what the pattern could mean. A response noting that such patterns can be associated with
conditions such as sleep apnea, and suggesting that the member discuss it with a clinician, is
precisely the kind of low-risk, high-value information the framework should enable.

Based on our understanding of functionality that is currently available in the market today, we do
not believe the Agency would take action against a foundation-model LLM offering context and
information on a user’s wearable readings (e.g., noting the possibility of a hypertension
diagnosis based on a blood pressure reading from a wearable device, and suggesting the user
seek advice from a healthcare professional). The Agency’s current wellness policy makes clear
that a wearable company cannot proactively notify a user about the same possibility because it
references a disease or condition. It leaves unclear, however, whether an LLM deployed by a
wellness company to help users interpret wellness-related data that produces similar output
would be treated in the same way. Similarly situated technologies with similar intended uses
should be treated similarly. LLMs that may interpret wearable data and provide interpretive
guidance to users on potential diseases and conditions, including guidance to consider
evaluation by a healthcare professional, should be held to the same standard by the
Agency, regardless of who deploys them.

Oura shares the Agency's view that GenAI will reshape the health technology landscape, and
we believe wearables will be one of the primary ways Americans experience its benefits. A
framework that anchors low-risk informational functions in existing enforcement discretion policy,
draws a clear boundary with the wellness guidance, and applies the same standard to all
deployers would protect users while allowing responsible companies to deliver that value. Thank
you for your consideration of these comments. Oura would welcome the opportunity to discuss
any of them in more detail.

Respectfully Submitted,

Ricky Bloomfield, MD
Chief Medical Officer, Oura