FDA Classification Rules for AI Medical Devices

 Posted on April 04, 2026 in AI Governance and Data Privacy

If you are developing or implementing an AI-powered medical device in 2026, understanding how the FDA classifies it is one of the first steps toward getting it to market or legally putting it into use. The classification your device gets determines how much oversight is required, what you need to submit to the FDA, and how long the process will take. Getting this wrong can cost you time and money, whether you are building the device or deploying it in a clinical setting. Our New York City AI governance lawyers can help you figure out where your device fits and what you need to do next.

How Does the FDA Define an AI Medical Device?

The FDA defines a medical device under the Federal Food, Drug, and Cosmetic Act, 21 U.S.C. § 321(h), as any instrument or software meant to diagnose, treat, cure, prevent, or reduce a disease or condition. Software that uses artificial intelligence or machine learning to do any of those things is called software as a medical device, or SaMD, and falls under FDA oversight.

Not every AI tool used in healthcare counts as a medical device. Software used for billing, scheduling, or general office tasks does not fall under the FDA's definition. However, software that looks at patient data to support a diagnosis, predict a health outcome, or suggest a treatment does. That kind of software has to be classified and regulated. This applies whether you built the software yourself or purchased it from a vendor to use in your practice or facility.

What Are the Three FDA Medical Device Classifications?

The FDA puts medical devices, including AI ones, into three classes based on how much risk they pose to patients. The higher the risk, the more review is required.

Class I: Lowest Risk

Class I devices have to follow basic rules like proper labeling, manufacturing standards, and FDA registration. Most Class I devices do not need premarket review. An example might be AI software that organizes medical images but does not draw any conclusions about them.

Class II: Moderate Risk

Class II devices need to follow additional controls, and most require a premarket notification submission called a 510(k). The manufacturer has to show that the new device is substantially equivalent to one already legally on the market. Many AI diagnostic support tools land here.

Class III: Highest Risk

Class III devices require the most review and generally need a Premarket Approval, called a PMA. This involves a full scientific review of clinical data to show the device is safe and effective. AI devices that make independent clinical decisions or that work to keep patients alive are most likely to end up in this class.

What Is the 510(k) Pathway and When Does It Apply to AI Medical Devices?

The 510(k) pathway is the most common route to market for Class II devices, including many AI tools. To use it, the manufacturer has to show their device is substantially equivalent to a predicate, which is a device already legally on the market that serves the same purpose.

For AI devices, finding the right predicate can be hard because the technology is new and the pool of already-cleared AI devices is still growing. The FDA has been clearing more AI and machine learning devices through this pathway in recent years, but putting together a strong, substantial equivalence argument is still one of the most critical parts of the process.

What Is De Novo Classification for AI Medical Devices and When Is It Used?

When a new device is low to moderate risk but has no predicate to compare it to, the manufacturer can ask for De Novo classification. This creates a new device type and sets the rules that will apply to it and to similar devices going forward. It takes longer than the 510(k) process but is faster than PMA. It is being used more and more often for new AI devices that do not fit into any existing category.

If a De Novo request is approved, the device becomes a predicate that other manufacturers can point to in their own 510(k) submissions. That means a successful De Novo can open a path for a whole category of similar AI tools.

What Do Businesses Using AI Medical Devices Need To Know About FDA Compliance?

Businesses that implement AI medical devices, such as hospitals, clinics, and health systems, are not off the hook when it comes to FDA compliance. If your organization customizes, modifies, or integrates an AI tool in a way that changes its intended use or performance, the FDA may consider your organization to be a manufacturer. This means you could take on regulatory responsibilities of your own.

Even without modification, implementing organizations need to make sure the devices they purchase and use are properly cleared or approved for their intended use. Using an AI tool for a purpose it was not cleared for can create serious legal and regulatory exposure, even if it seems like a logical extension of its capabilities. Before deploying any AI medical device, it is worth confirming that the device is cleared for exactly the way you plan to use it.

How Does the FDA Handle AI Medical Devices That Learn and Change Over Time?

Traditional medical devices do not change after they are cleared. AI devices that use machine learning can evolve as they process more data. That means the device a company cleared today may have meaningfully different capabilities in just a few months from now.

The FDA has addressed this with guidance on predetermined change control plans, called PCCPs. A PCCP lets a manufacturer describe upfront what kinds of changes the algorithm might go through and how those changes will be checked before they go live. This way, the FDA can review the plan at the start rather than requiring a new submission every time the algorithm updates. For AI devices with adaptive algorithms, including a PCCP in the initial submission is becoming more and more important.

Contact Our New York City AI Medical Device Compliance Attorneys Today

FDA classification for AI medical devices sits at the crossroads of technology law, regulatory compliance, and product liability. Whether you are building an AI medical device or putting one to work in your organization, the rules apply to you.

At CO Health Advisory, we work with early and mid-stage startups and companies on a budget. We offer general and outside counsel services so we can be your ongoing legal resource as your product and regulatory strategy develop. Call 212-739-0611 to talk to our experienced New York City AI governance lawyers.