Skip to main content
☩ AI Health Guide
Menu

FDA-Cleared AI Health Tools 2026: What Clearance Actually Means

Most AI health products marketed today are not FDA-cleared, and most do not need to be. A small set of clinically critical AI tools — primarily in radiology, pathology, and cardiac imaging — have gone through 510(k) clearance, De Novo authorization, or Breakthrough Device Designation. This guide covers the leaders, what their clearances actually authorize, and how to interpret regulatory status when evaluating AI health tools.

A

AI Health Apps Editorial Team

Editorial Team, AI Health Guide

Updated

Reviewed against our methodology

About the editorial team  ·  Report an error

What FDA Clearance Means (and Doesn't)

FDA clearance for an AI health tool typically falls into one of several pathways. 510(k) clearance means the FDA has determined the device is "substantially equivalent" to a legally marketed predicate device for a specified intended use. It is the most common pathway and the one most AI imaging tools have used. De Novo authorization is for novel low-to-moderate-risk devices without an existing predicate — Viz.ai received the first AI triage software De Novo clearance in 2018. Premarket Approval (PMA) is the highest bar, used for high-risk devices, and is rare in current AI health. Breakthrough Device Designation is not a clearance — it is a status that gives expedited FDA engagement for devices addressing serious or life-threatening conditions.

FDA clearance authorizes a specific intended use, not general capability. A clearance for "AI-powered triage notification of large vessel occlusion in suspected stroke" does not authorize the same system to triage hemorrhagic stroke unless that indication is also cleared. When evaluating AI health tools, always read the actual indication for use, not the marketing summary.

Clearance also does not equal clinical efficacy in your specific patient population. FDA review evaluates whether the device performs as claimed, not whether deploying it will improve outcomes in your hospital. Local validation is still essential.

The Leaders in 2026

Viz.ai — First AI Triage De Novo and a Growing Set of Clearances

Viz.ai's ContaCT received an FDA De Novo grant in 2018 (DEN170073) for the detection of suspected large vessel occlusion in stroke; FDA said the decision created a new regulatory classification for computer-aided triage software, so later triage tools could use the 510(k) route. FDA's device database now lists 10 FDA 510(k) clearances and 2 De Novo grants for Viz.ai (openFDA data as of 19 October 2026), across stroke (LVO), intracranial hemorrhage, aneurysm, aortic disease, subdural hemorrhage, and cardiac analysis. Viz.ai also established the Medicare NTAP reimbursement pathway for AI triage software, which is a meaningful precedent for AI device economics.

Read the Viz.ai review.

Aidoc — Comprehensive Foundation Model Clearance (January 2026)

Aidoc received FDA 510(k) clearance (K252970) in January 2026 for a comprehensive body CT triage foundation model, covering 14 acute conditions (11 new plus 3 existing) from a single CT scan; Aidoc reports 97% mean sensitivity and 98% mean specificity for the 11 new indications. Aidoc calls it the first FDA clearance of a double-digit set of acute indications from a single foundation model; the FDA record itself does not rank it. Aidoc is deployed in 1,600+ hospitals globally including 7 of the top 10 US hospitals.

Read the Aidoc review.

Tempus AI — Multiple Cleared Genomic and Cardiac AI Products

Tempus AI holds FDA authorizations spanning genomic sequencing and cardiac AI. Its DNA next-generation sequencing panel, xT CDx, is FDA-approved through premarket approval (PMA P210011, April 2023), not a 510(k) clearance. Its 510(k) clearances include Tempus xR IVD (RNA NGS, cleared September 2025, K241868), Tempus Pixel (cardiac MRI AI, K252539), and Tempus ECG-Low EF (detection of reduced ejection fraction from a standard ECG, K250119). Few vendors we cover hold cleared genomic tests and cleared imaging and ECG AI under one company.

Read the Tempus AI review.

PathAI — First DDT Biomarker Qualification for AI Pathology

PathAI holds three kinds of FDA recognition. AISight Dx received FDA 510(k) clearance (K243391) in June 2025 as software for viewing and managing digital images of surgical pathology slides, an aid to the pathologist for primary diagnosis. Its Predetermined Change Control Plan lets PathAI add FDA-cleared displays, scanners, file formats and browsers without a new submission (510(k) summary); it does not cover AI model updates. AIM-MASH AI Assist, qualified by the FDA under its earlier name AIM-NASH, received FDA Drug Development Tool Biomarker Qualification in December 2025; FDA called it the first AI drug development tool it has qualified, for use in MASH clinical trials. PathAssist Derm received Breakthrough Device Designation in March 2026, per PathAI.

The DDT qualification matters for drug developers: within its qualified context of use, sponsors can use the tool in IND, NDA and BLA submissions without resubmitting the tool's validation (per PathAI). It is a drug-development qualification, not a device clearance for clinical care.

Read the PathAI review.

What We Did NOT Include

This list intentionally excludes AI tools that do not have FDA device clearance. That includes most ambient AI medical scribes (Abridge, Suki, Freed, DeepScribe, Nuance DAX Copilot), most AI EHR features (athenahealth AI, DrChrono AI), all AI mental health chatbots that previously needed clearance (Wysa holds Breakthrough Device Designation but not full clearance), and consumer symptom checkers (such as Ada Health). These products are not FDA-cleared because they generally fall outside the FDA device definition — ambient documentation aids and clinical decision support tools that meet the 21st Century Cures Act non-device CDS criteria are exempt from FDA device regulation.

Lack of FDA clearance does not mean a product is unsafe or unvalidated. It generally means the product does not make a diagnostic, therapeutic, or treatment claim that requires clearance. For ambient AI scribes, the "documentation aid" framing is what keeps them outside the FDA device pathway, and the FDA's mid-2025 guidance clarifying AI scribe regulatory classification confirmed this.

How to Read Regulatory Status as a Buyer

Three questions to ask any AI health vendor about FDA status:

  1. What is the exact intended use cleared by FDA? Marketing language often expands beyond the cleared indication. Ask for the K-number and read the indication for use statement directly.
  2. Do you have a Predetermined Change Control Plan (PCCP), and what does it cover? A PCCP lets the vendor make the specific changes it describes without a new FDA submission. Read its scope: PathAI's AISight Dx PCCP, for example, covers adding cleared displays, scanners, file formats and browsers, not changes to an AI model.
  3. What is your post-market surveillance and performance monitoring approach? AI models drift. Vendors with mature governance can describe their monitoring program clearly. Aidoc's continuous performance monitoring is a useful reference standard.

The Bottom Line

FDA-cleared AI health tools cluster in radiology, pathology, and cardiac imaging because those are the categories where AI is making decisions about specific clinical indications that require regulatory oversight. Viz.ai and Aidoc are the radiology triage vendors covered here, Tempus AI spans genomic and cardiac AI, and PathAI holds both a 510(k) clearance and an FDA drug development tool qualification in pathology. For health systems building an AI strategy, these are the safe starting points where the regulatory question is settled. For everything else, the regulatory framing may be accurate but is not a substitute for local validation.

Frequently asked questions

What does "FDA-cleared" actually mean for an AI health tool?
FDA "clearance" via the 510(k) pathway means the FDA has reviewed and accepted the device as substantially equivalent to a legally marketed predicate device for a defined indication. It is not the same as FDA "approval" (granted via PMA for higher-risk devices). For AI/ML tools, clearance typically applies to a specific clinical indication (e.g. detecting intracranial haemorrhage on non-contrast CT) — using the tool outside that indication is off-label.
Which AI radiology tool has the most FDA clearances?
Of the two radiology triage vendors we cover, Aidoc has more on record: FDA's 510(k) database lists 35 clearances for Aidoc, and 10 510(k) clearances plus 2 De Novo grants for Viz.ai (openFDA data as of 19 October 2026). We have not counted every radiology AI vendor. Aidoc also received FDA clearance in January 2026 for its comprehensive foundation model AI covering 14 acute conditions on a single CT scan.
Is FDA-cleared AI Medicare-reimbursable?
Some FDA-cleared AI tools have CMS reimbursement codes — Viz.ai has had a New Technology Add-on Payment for stroke triage since 2020, and Tempus AI has reimbursable genomic-AI offerings. Most FDA-cleared AI tools are still pursuing reimbursement and are paid for by hospital capital budgets, not per-use Medicare codes.
Are AI medical scribes FDA-regulated?
Most AI medical scribes (Abridge, Nuance DAX, Suki, Freed, DeepScribe) operate under FDA guidance carve-outs as Clinical Decision Support tools and are not classified as medical devices. The FDA released guidance in mid-2025 clarifying when an AI scribe crosses the threshold into device regulation — primarily when the AI is making autonomous diagnostic or treatment decisions, not just transcribing.
Has the FDA cleared any AI for pathology?
PathAI received the first FDA qualification for an AI pathology tool in clinical trials in late 2025 — specifically AIM-NASH (the FDA's name; PathAI now sells it as AIM-MASH AI Assist), an algorithm for assessing metabolic dysfunction–associated steatohepatitis. This qualification is for use in pharmaceutical clinical trials, not for routine clinical practice. Separately, PathAI's AISight Dx has a 510(k) clearance (K243391, June 2025) as software for viewing and managing digital slide images, an aid to the pathologist for primary diagnosis; it is not an AI diagnostic algorithm.
Can I trust AI tools that are not FDA-cleared?
FDA clearance is one signal of validation, not the only one. Some legitimately useful AI tools (consumer mental-health apps, clinical-decision-support tools that fall under FDA carve-outs) are intentionally unregulated. Look for: peer-reviewed clinical validation, transparency about the training data and validation cohort, third-party benchmarking (e.g. KLAS for B2B health tech), and a clear statement of intended use.
For Clinicians: Freed Try Freed ↗