FabolaAI medical devices

AI-enabled medical devices

A model performing well is not the same thing as a medical device being adequately validated.

Fabola works where intended purpose, datasets, clinical evidence, model performance, lifecycle change and MDR/IVDR–AI Act requirements meet. For an AI-enabled medical device, the technical, clinical and regulatory parts of the company need to connect: choices about data, model updates and intended use affect all three.

Two people reviewing results on a laptop

Where the questions meet

Most difficult AI regulatory questions sit between disciplines. These are the areas where they usually come together.

Intended purpose and claims

What the AI is claimed to do decides how the product qualifies, how it is classified and what evidence it needs. A precise intended purpose is the starting point for everything else.

Datasets

Training, test and validation data need to be documented, kept appropriately independent and representative of the population the device is intended for.

Clinical evidence and model performance

Performance metrics only count as evidence when the metrics, reference standard, population and acceptance criteria are tied to the intended medical use and its clinical risks.

Lifecycle change

Models change. Which changes are planned, how they are controlled and when a change needs new evidence or a new assessment should be settled before launch, not after.

MDR / IVDR and the EU AI Act

An AI system that is, or is a safety component of, a medical device or IVD requiring Notified Body assessment is treated as high-risk under the AI Act. The two sets of requirements overlap, and they should be met through one technical documentation, not two parallel programmes.

Human and AI together

How clinicians or users interact with the output, and what happens when the model is wrong, belong in the risk analysis and the validation, not only in the user interface.

Questions we work on

  • What exactly are we claiming the AI does?
  • Is the validation dataset representative of the intended population?
  • How should subgroup performance and bias be evaluated?
  • Which performance metrics matter clinically and regulatorily?
  • How should model error be linked to clinical risk?
  • What happens when the model changes?
  • How do MDR or IVDR and the EU AI Act interact for our product?
  • Does using a large language model change how the product is regulated?

Services for AI-enabled devices

AI validation and evidence

How do we demonstrate that our AI-enabled medical device works reliably and safely for its intended medical use?

EU AI Act + MDR / IVDR

How does the EU AI Act interact with the MDR or IVDR for our product?

Clinical and performance evidence

What evidence do we actually need to support our medical device or IVD?

View all services

Primary regulatory sources

Last reviewed: October 2026

Working on an AI-enabled device?

Tell us what the model does, where the product stands and the regulatory question you need answered.

We will point you to the right piece of work, or tell you if the project needs a broader scope.

Discuss your AI medical device

The first conversation is free and carries no commitment.