I’m entering words from my browser into my code and Copilot guesses the word before I begin to type it. The only way it could know what I want to type 75% of the time, I think, is if it can see the website in my browser. Is this right?
It’s Microsoft. Assume surveillance.
I’ve seen this thread somewhere else before.
Same thing different context.
And I seem just like before that it’s because the user is entering what are very predictable words or phrases and they are just not putting two and two together.
I wondered this too when I was using the Jetbrains ai a while back, but I came to the conclusion that it’s probably just trained on code that’s on the internet, so you’re likely to type something that’s literally in its trainingdata when you’re copying something from the web.
What I will say needs to be fact check because I never dig much on it since I’m not using w11 so I can be wrong. There is a feature called recall which take regular snapshot on w11.
A quick search about gives different results but it is apparently a opt in feature, so unless you enabled it, it shouldn’t.
If you think it does this from your browser search, try using a second device which isn’t windows based to browse the net and see if its still guesses right
Recall is only on copilot+ PCs, not all windows 11 PCs
always assume you’re compromised to some degree
The only way I know Forsthaus to work, you would need to use something like the playwright mcp, and use it to let the LLM view what you are browsing. Not sure that even has a GUI.
I was wondering the same thing.
Weird privacy things like this happen to me as well. I’m just putting it down to the placebo effect, I don’t think it really can see your browser unless microsoft is exploiting some security vulnerability in your os.
