FDA-Authorized AI Medical Devices 2026: What the List Really Means

Last updated: October 4, 2026

AI is already part of well over a thousand medical devices marketed in the United States. Headlines call them «FDA-approved AI,» but that label hides a lot. The FDA does publish a list of AI-enabled devices, yet it is not a register of «approved AI.» It includes devices that were cleared, granted or approved through different pathways, and the difference matters.

This article explains how large the list is in 2026, which specialties dominate it, and what authorization does and does not tell you. It builds on our articles on AI symptom checkers and AI in radiology.

How We Researched This Article

This is a desk-research article. We did not test any device. We:

• Used the FDA’s own pages as the primary source for counts, pathways and policy dates.
• Prioritized peer-reviewed analyses of the FDA list over press summaries, and checked numbers against the original pages as of October 2026.
• Noted where a figure comes from trade-press reporting on an FDA update that we could not open directly.

How Many AI Medical Devices Has the FDA Authorized?

The FDA states that its public list includes more than 1,600 AI-enabled medical devices as of September 2026. Treat that as a minimum: the list changes with each update, and the FDA says it is not comprehensive. It includes devices identified mainly through AI-related terms in their public summaries, so some devices may be missing.

FDA list snapshot

What the Evidence Shows

«FDA Approved» or «FDA Authorized»?

This is the most common source of confusion. Most devices on the FDA’s AI-enabled list were cleared through 510(k), rather than approved through the PMA pathway.

A 510(k) decision is based primarily on substantial equivalence to a legally marketed predicate. Clinical or performance data may be submitted or requested depending on the device and the regulatory question, but clearance should not automatically be interpreted as evidence that the FDA independently conducted a new clinical trial. That does not make clearance meaningless, but it answers a different question from the one many readers assume.

Why Does Radiology Dominate?

In the update covering authorizations through the end of 2025, 1,104 of 1,451 devices, about 76%, were radiology. In the fourth quarter of 2025 alone, 72 devices were cleared and 55 were radiology, according to trade-press coverage of the FDA update. Imaging produces large amounts of standardized digital data (CT, MRI, X-ray, mammography, ultrasound) that suits machine learning. Radiology tools may highlight suspicious findings, measure structures, improve reconstruction or prioritize urgent studies. Some assist a radiologist while others automate part of a workflow (see AI in Radiology in 2026).

This concentration does not mean radiology is the only medical specialty using AI; it means radiology has historically accounted for a disproportionately large share of the devices identified in the FDA’s list.

Two Early Examples of FDA-Authorized AI

These two show that «medical AI» is not one technology: purpose, pathway and degree of automation vary widely.

What FDA Authorization Does Not Tell You

Authorization shows that a device met the requirements of its pathway for its intended use. It does not automatically show that:

• the product is the most accurate AI available;
• it improves patient survival or outcomes;
• it works equally well in every hospital, scanner and patient group;
• it outperforms an experienced clinician;
• the manufacturer’s wider marketing claims are covered.

The 2025 analyses, which examined devices on the FDA’s AI-enabled list rather than only those formally approved, help explain why authorization counts should not be treated as outcome evidence. Only 2.4% of reported clinical studies were randomized, fewer than one third included sex-specific data, and 4.8% of devices had been recalled by the time of analysis. The authors call for ongoing monitoring and re-evaluation after authorization.

What About Generative AI?

On August 18, 2026, the FDA published a discussion paper, Considerations for the Regulation of Generative AI-Enabled Medical Devices, covering risk assessment, premarket evaluation and postmarket monitoring. Comments are open until October 19, 2026 (docket FDA-2026-N-7874). The paper is for discussion only and is not draft or final guidance. The publication of the discussion paper does not mean that general-purpose consumer chatbots are FDA-authorized medical devices: status depends on a product’s intended use, not on the technology behind it.

How to Check an AI Medical Device Yourself

1. Search the FDA’s AI-enabled device list using the exact company and product name. If it is missing, check the FDA’s device databases too, since the list is not comprehensive.
2. Note the pathway (510(k), De Novo or PMA) and the decision date.
3. Read the indications for use, which define what the device may do and for whom.
4. Review the testing and patient populations in the summary documents.
5. Compare all of this with the manufacturer’s marketing claims.
6. Check whether the authorization covers the exact version and intended use you are evaluating.

Limitations of This Review

• Counts differ slightly between analyses because they use different download dates (for example 1,430 versus 1,451 devices through 2025).
• The end-2025 radiology and Q4 2025 figures come from trade-press reporting on the FDA update, which we could not open directly.
• Public summaries do not show everything the FDA reviewed, so missing evidence in them is not proof of absence.
• The FDA list is not comprehensive, and totals change with every update.

Frequently Asked Questions

Final Thoughts

The FDA list shows how fast AI has entered regulated medicine, and how concentrated it is in radiology. But the number of authorizations is not the number of clinically proven technologies.

Related Reading

Sources

Not every figure in this article comes directly from the FDA. Counts and policy dates come from the FDA’s own pages (sources 1 and 2); study-level findings come from peer-reviewed analyses (sources 3 to 5); the end-2025, Q4 2025 and March 2026 radiology figures come from trade-press reporting (source 6); and details of the 2018 authorizations come from legal and trade coverage (source 7).

1. U.S. FDA. Artificial Intelligence-Enabled Medical Devices (content current as of September 22, 2026) and the FDA AI-Enabled Medical Devices List page.
2. U.S. FDA. Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion Paper and Request for Feedback, August 18, 2026.
3. Sivakumar R, Lue B, Kundu S. FDA Approval of AI and ML Devices in Radiology: A Systematic Review. JAMA Netw Open 2025;8(11): e2542338.
4. Windecker D et al. Generalizability of FDA-Approved AI-Enabled Medical Devices for Clinical Use. JAMA Netw Open 2025;8(4): e258052.
5. Clinical Specialty Expansion of AI-Enabled and ML-Enabled Medical Devices Authorized by the FDA From 1995 to 2025. J Med Internet Res 2026;28: e103040.
6. The Imaging Wire. Numbers from the FDA show radiology is maintaining its lead, March 2026 (could not be opened directly; figures cross-checked against other trade-press reports).
7. IDx-DR De Novo authorization, April 2018: Jones Day legal alert and MobiHealthNews coverage; Viz.ai Contact De Novo, February 2018: MobiHealthNews FDA roundup.