October 19, 2026, is the deadline to submit comments on Regulations.gov under docket FDA-2026-N-7874. FDA is seeking early feedback on generative AI medical device regulation through a discussion paper rather than proposing policy changes.
Key Takeaways
- FDA announced the discussion paper on August 18, 2026. It is neither draft nor final guidance.
- The possible risk framework considers independent device activity and the severity of harm from relying on incorrect output.
- Feedback covers risk assessment, premarket evaluation, postmarket monitoring, and related issues. Comments are due October 19, 2026.
What FDA is opening for discussion
FDA’s discussion paper is a request for early stakeholder input on generative AI-enabled medical devices. Announced on August 18, 2026, it addresses how to assess risk and evaluate and monitor these devices. FDA explicitly says the paper does not propose or implement policy changes.
The document’s full title is Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion Paper and Request for Feedback. Its docket is FDA-2026-N-7874. Keep that identifier with the document when sharing it internally, particularly if your team is preparing comments.
A discussion paper invites input; it does not establish the agency’s regulatory expectations. FDA says this document is neither draft nor final guidance and is intended to advance stakeholder discussion. Describe it that way in a procurement memo or product review, rather than treating the framework as an adopted requirement.
For your organization, start by separating what FDA is exploring from questions about your own device. Bring questions about applicable requirements to counsel or the regulator. Use our guide to FDA and peer-reviewed sources alongside the primary document when reviewing the claims you plan to circulate.
Understanding the two axes of device risk
The paper presents a possible framework that places device activity on one axis and consequences on the other. Activity concerns how independently a function directs or takes action. Consequences concern the severity of harm from relying on an incorrect output, with increasing risk toward the grid’s upper-right corner.
In plain language: activity describes independence of action; consequences describe harm if the output is wrong. Read those dimensions together when discussing a software function. The paper contrasts a function operating under continuous healthcare professional supervision with one operating autonomously.
| Element | What the paper describes | Example or relationship |
|---|---|---|
| Activity | How independently a function directs or takes action | A non-directive risk score versus a strongly worded patient-facing recommendation |
| Consequences | Severity of harm from relying on incorrect output | An incorrect over-the-counter remedy suggestion versus an incorrect insulin-dose adjustment recommendation |
| Risk gradient | Increasing risk across the combined dimensions | From the lower-left toward the upper-right of the grid |
FDA uses those examples to distinguish activity and potential consequences. They are illustrations within a discussion framework. Avoid turning the grid into a claimed compliance score or assigning a product a regulatory status from it.
For an internal review, describe the function before debating its position on the grid. Who supervises its operation? Does it present information or direct action? What harm could follow reliance on incorrect output? Frame your answers as observations for discussion, with questions for counsel where needed.
Change control for foundation models and agentic systems
FDA describes foundation models as broadly applicable AI models trained on large, typically unlabeled datasets using substantial computational resources. They can serve as bases for further fine-tuning. Their capabilities can also extend to contexts beyond those for which they were originally developed and trained.
Agentic AI systems, in the paper’s terminology, use generative AI to plan and execute tasks autonomously across multiple steps. They may use external tools or take actions through a sequence. When discussing such a function, connect that definition to the activity axis: describe the actions and degree of independence.
The paper also addresses how third-party foundation models complicate change control. It discusses a Foundation Model Device Master File concept in connection with third-party model changes. Treat that concept as a discussion topic, consistent with the paper’s stated purpose.
Your team could organize feedback around the relationship between the model and the device function. Describe which activities depend on the model and what questions a model change raises for your review. Keep proposed solutions clearly labeled as your suggestions, without presenting the master file concept as an established obligation.
Evaluation and monitoring in the broader FDA context
The paper seeks feedback on premarket evaluation and postmarket monitoring as well as risk assessment. Its scope therefore includes discussion of evaluation before marketing and monitoring afterward. Read those topics alongside the change-control discussion when deciding which parts warrant comments from your organization.
For a focused internal conversation, separate questions about evaluation from questions about ongoing monitoring. Then identify where third-party model changes connect the discussions. This is an editorial way to organize your response, rather than a submission structure prescribed by the paper.
FDA’s AI device page states that it has authorized over 1,600 AI-enabled medical devices for marketing in the United States as of September 2026. The page is marked current as of September 22, 2026. Preserve that date when citing the total, and retain FDA’s broader AI-enabled wording.
You can also consult our earlier FDA-authorized AI device tracker coverage for related reading. For the figure discussed here, use FDA’s dated statement directly. Avoid describing that total as a count of generative AI devices.
The same FDA page lists Artificial Intelligence-Enabled Device Software Functions: Lifecycle Management and Marketing Submission Recommendations as draft guidance. It also links to the generative AI discussion paper. Keep these document labels distinct in your team’s notes: the guidance remains listed as draft, while the discussion paper solicits early input.
Submitting comments by October 19, 2026
FDA directs public comments to Regulations.gov under docket FDA-2026-N-7874, with a deadline of October 19, 2026. The discussion-paper landing page identifies both the docket and the deadline. Use that page as your starting reference when preparing a response.
Read the paper, select the topic your team wants to address, and draft comments around that issue. You might focus on the activity axis, consequences of incorrect output, evaluation, monitoring, or third-party model changes. Label your own examples and proposals clearly.
Before submitting, check the docket identifier against FDA’s landing page. For questions about your organization’s circumstances or regulatory obligations, consult counsel or the regulator. The task here is to contribute to the discussion FDA has opened.
Frequently Asked Questions
What did FDA announce on August 18, 2026?
FDA announced a discussion paper on generative AI-enabled medical devices under docket FDA-2026-N-7874, seeking early stakeholder feedback.
Is the discussion paper FDA policy or regulation?
No. FDA says the paper is neither draft nor final guidance and does not propose or implement policy changes.
What is the two-axis framework?
It is a possible risk framework combining independent device activity with the severity of harm from relying on incorrect output. Risk increases toward the upper-right of the grid.
How and by when can you comment?
Submit comments on Regulations.gov under docket FDA-2026-N-7874 by October 19, 2026.
Sources
- FDA seeks public feedback on generative AI-enabled medical devices. FDA, August 18, 2026.
- Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion Paper and Request for Feedback. FDA, announced August 18, 2026.
- Discussion paper and request for feedback: comment information. FDA, comment deadline October 19, 2026.
- Artificial Intelligence-Enabled Medical Devices. FDA, content current September 22, 2026.
This article is for general information and is not legal, regulatory or medical advice.