If it’s using a local model like it says I think this is fine:

We’re looking at how we can use local, on-device AI models – i.e., more private – to enhance your browsing experience further. One feature we’re starting with next quarter is AI-generated alt-text for images inserted into PDFs, which makes it more accessible to visually impaired users and people with learning disabilities.

Did you read more than just the three words naming the feature? Their use case is actually smart, and could potentially help users a lot.

The recent addition of local in-browser website translation is an awesome feature I’ve missed for many years. The only alternatives I’ve found previously were either paid or Google Translate plugins. This translation feature is an example of an AI feature they’ve added.