Skip to Content
Artificial intelligence

The FDA wants to regulate machine learning in health care

April 4, 2019

The US Food and Drug Administration has announced that it is preparing to regulate AI systems that can update and improve themselves as they gorge on more training data.

The announcement: The agency released a white paper proposing a regulatory framework to decide how medical products that use AI should seek approval before they can go on the market. It is the biggest step the FDA has taken to date toward formalizing oversight of products that use machine learning (ML).

The challenge: Machine-learning systems are tricky to regulate because they can continuously update and improve their performance through new training data. In instances where the FDA has approved ML-based medical software before, it has required the algorithms to be “frozen” before commercial deployment and to go through a reapproval process when they are changed. But the agency recognizes that this process means we lose out on the benefits of the technology. New rules might allow ML systems to evolve to a certain extent after initial approval while the risks of that evolution are managed.

The framework: When seeking approval for a given product, providers of ML medical devices would also have the option of submitting a plan for planned software modifications, including model retraining. Additionally, the FDA would evaluate the quality of the provider’s product development process to ensure, for example, that its system is verified through clinical trials.

Next steps: The FDA is now collecting feedback on the document. The paper “demonstrates careful forethought about the field,” Eric Topol, an expert at the intersection of deep learning and health care, told Stat.

For more on the world of AI, sign up to our Webby-nominated newsletter The Algorithm here.

Deep Dive

Artificial intelligence

Why Meta’s latest large language model survived only three days online

Galactica was supposed to help scientists. Instead, it mindlessly spat out biased and incorrect nonsense.

DeepMind’s game-playing AI has beaten a 50-year-old record in computer science

The new version of AlphaZero discovered a faster way to do matrix multiplication, a core problem in computing that affects thousands of everyday computer tasks.

A bot that watched 70,000 hours of Minecraft could unlock AI’s next big thing

Online videos are a vast and untapped source of training data—and OpenAI says it has a new way to use it.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.