Why is AI visibility in the browser layer so hard?
we blocked chatgpt and a few others at the network level months ago, but most AI usage just moved into the browser. trying to get visibility there and running into the same issues over and over:
- extensions and sidebars (copilot, claude, random plugins) run client-side or through approved domains, so there’s nothing obvious to block
- network logs don’t show much since traffic blends in with normal SaaS usage
- CASB catches some standalone tools but misses local extensions and embedded features
- chrome enterprise policies help a bit but don’t cover everything, plus users complain about performance
- no clear way to see what’s being pasted into prompts or what data is leaving
- devtools show some calls but not something you can realistically monitor across an org
at this point we know usage is happening just don’t have a clean way to see it.
anyone figured out browser-layer visibility without killing performance or rolling out full endpoint agents. what are you actually running for this?