Microsoft Edge says it may download and run generative AI models on your device

Microsoft Edge is updating its On-device AI setting in Canary to explain local AI model downloads more clearly. The browser now says it may download and run generative AI models on your device for more private, local processing.

The new description reads:

“Allow Edge to download and run generative AI models on your device for more private, local processing. If off, some features and sites may not work as expected.”

The earlier version we covered said Edge could use AI models that run directly on your device. It read: “To power supported features, Microsoft Edge can use AI models that run directly on your device. When this is off, these features might not work.”

The new wording explicitly mentions downloading generative AI models and running them locally.

Microsoft hasn’t listed in the description which features or sites rely on these models.

Edge already supports on-device AI through its Prompt and Writing Assistance APIs. These APIs use a small language model that Edge downloads when a website first calls an API that needs it. Edge then shares the downloaded model across websites that use these APIs.

Downloading a model also takes up storage space and uses bandwidth. Edge doesn’t necessarily download one as soon as you enable the setting. A website or feature must first request an API that needs the model, and Edge must meet the download requirements.

Microsoft hasn’t announced any changes to how On-device AI works. The updated description simply gives users a clearer explanation of local AI model downloads.

Chrome has also recently updated its On-device AI setting and help page to give users more clarity about how local AI models work.