I think it a distinction without a difference. To make it more obvious imagine it was one of those AI assistant devices that records your conversations so you can recall them later. Plainly obvious that accessing this data for any purpose other than servicing user requests is morally equivalent to easedropping on a person's conversations in the most traditional sense.

If the company sends your conversation data to Facebook that's bad and certainly a privacy violation but at this point nothing has actually been done with the data yet. Then Facebook accesses the data and folds it into their advertising signals; they have now actually looked at the data and acted on the information within. And that to me is easedropping.

To extend your analogy further, what if instead of an AI assistant, it was your friend who listened to your secret, and instead of him sending your data to facebook, he told that to google (eg. "hey gemini, my friend has hemorrhoids..."). Suppose further that google uses Gemini queries for advertising purposes (eg. upranking ad results for hemorrhoid creams). Should gemini be on the hook for this breach of trust? What if, instead of a big evil tech company, it was the owner of a local corner shop, who uses this factoid to inform his purchasing decisions?

I disagree - the blame lies with the people who sent that data to Facebook knowing it was sensitive. Whrhther meta use it for advertising or not is irrelevant.

By that logic, if I listen in on your conversations but don’t do anything about it I’m not eavesdropping?