Chrome now shows a stronger warning when an extension wants broad page access and markets itself as an AI sidekick.
Google is not banning AI browser extensions. It is making the scary permission dialog harder to click past without reading. Chrome will now surface a clearer warning when an add-on asks to see almost every site and also sells itself as an AI helper. The install button remains. The friction is the point.
Why this warning exists
A lot of “AI sidekick” extensions need host permissions that amount to reading the page you are on. Some of that is legitimate: summarizing a doc in a tab really does require seeing the tab. Some of it is lazy product design dressed up as magic. Google said the heuristic looks at permission breadth and store listing language, not at whether the model behind the button is any good. That is the right layer for a browser to police. Chrome cannot grade answer quality. It can flag “this thing wants the keys to every site.”
The new copy names the host permissions in plainer language and points people at Chrome’s guidance on page-reading extensions. Power users who already scrutinize permissions will shrug. The audience is everyone else: the install that happens because a YouTube video said the extension writes emails for you.
What users and IT should do
Before you approve anything that rewrites forms or “reads the web for you,” open the details and ask which sites it can touch. Prefer extensions that request access only when you click them. If your company already manages Chrome, keep allowlisting the tools you vetted and leave the consumer warning alone for staff on unmanaged machines.
None of this replaces basic hygiene. An extension with narrow permissions can still be malicious. An extension with wide permissions can still be careful. The update simply makes the wide-permission AI pitch harder to treat as a free candy bowl. That is a small, useful change in a storefront full of confident buttons.