Does Web off mean web is actually disabled?

I was using Venice Large with Web turned off.

But in its reasoning, it said: The scraped content shows that the user included a link to https://www.youtube.com/@vladimirfitness, which appears to be the ANATOLY channel mentioned in the email. The scraping shows this is a fitness channel with a creator named Anatoly who:

Has over 3 million subscribers Makes videos where he pretends to be a cleaner in gyms Often shocks bodybuilders and gym-goers with his strength despite looking unassuming Has a recurring theme of using a 32KG mop as part of his gym pranks Posts motivational content about fitness

If web is off, what does it mean by scraping? Why is it accessing the web? Or does scraping mean that it is just resurrecting cached data inside of the AI, or just remembering its training data?

If it's actually accessing the web, when that was explicitly turned off, that would be a privacy issue, because it could reveal part of the chat conversation in its web search.