FDA AI/ML Device Predicate Networks

Explore how FDA-cleared AI devices rely on earlier predicates across specialties and pathways

Devices
Links

Predicate Provenance

Same-specialty AI
Other-specialty AI
Non-AI device
No predicate

FDA Pathway

510(k)
De Novo
Premarket
Pinch to zoom · Drag to pan · Click for details

Device Details

Click a device node to view details

About This Dashboard

This interactive tool maps the regulatory relationships between FDA-approved AI/ML-enabled medical devices. It reveals how each device gained market clearance and which earlier devices it relied upon — making the often opaque regulatory pathway visible and explorable.

Why does this matter?

The vast majority of AI medical devices reach the market through the 510(k) pathway, which requires manufacturers to demonstrate that their device is “substantially equivalent” to a previously approved device called a predicate. This means a new AI diagnostic tool might be cleared not on the basis of its own clinical trials, but because it is deemed similar enough to an older device — which may itself not even be an AI device. Understanding these chains of equivalence is critical for clinicians, researchers, regulators, and patients who want to know: what evidence actually underpins this device’s approval?

How the FDA approves AI devices

The FDA uses three main regulatory pathways:

  1. 510(k) Premarket Clearance — the most common route (~97% of AI devices). The manufacturer demonstrates substantial equivalence to a predicate device. Does not necessarily require new clinical trials.
  2. De Novo Classification — for novel, low-to-moderate risk devices with no suitable predicate. Requires more evidence than 510(k) but less than PMA. Once granted, the device itself becomes a predicate for future 510(k) submissions.
  3. Premarket Approval (PMA) — the most rigorous pathway, reserved for high-risk devices. Requires clinical evidence of safety and effectiveness. Rare for AI devices.

How to read the graph

How to use the dashboard

Key observations

A significant proportion of AI devices cite non-AI predicates (terracotta dots), meaning their regulatory clearance is based on equivalence to devices that predate the AI era entirely. The tree structures reveal how a single early approval — such as a De Novo classification — can become the regulatory foundation for dozens of subsequent devices across multiple companies and clinical applications.

We aim to regularly update these graphs with more specialties, features, and up-to-date devices. The FDA AI/ML database is periodically updated.

Hover over terms below for definitions:

Predicate Device
A previously approved device cited as the basis for a new 510(k) submission. The new device must be “substantially equivalent” to its predicate in intended use and technology.
Predicate Network
The full chain of devices connected through predicate relationships. Clicking a device highlights every device in its network — both ancestors (what it was cleared against) and descendants (what was cleared against it).
Predicate Provenance
Classifies where a device's predicate sits: within the AI/ML database (AI device), outside it (non-AI device), or absent entirely (De Novo / no predicate). This reveals whether AI devices are being cleared against other AI devices or against older, non-AI technology.
510(k)
The most common FDA pathway for AI devices. Based on demonstrating “substantial equivalence” to a predicate device. Does not inherently require new clinical trials, though the FDA may request additional data.
De Novo
A pathway for novel devices with no suitable predicate. Requires a risk-based assessment. Once approved, the device creates a new regulatory category and can serve as a predicate for future 510(k) submissions.
Submission Number
A unique FDA identifier for each device application. 510(k) numbers start with “K” (e.g. K231335), De Novo with “DEN” (e.g. DEN170073), and PMA with “P”.
Substantial Equivalence
The legal standard for 510(k) clearance. A device is substantially equivalent if it has the same intended use and either the same technological characteristics, or different characteristics that do not raise new safety or effectiveness questions.