Honestly, I could see this being a viable method of sending our info out. But it would have to be an LLM trained on very well curated input data. And, of course, there's the little issue of "how the fuck do the people receiving it understand what it is saying and how does it understand what they are saying". I'm sure some very clever people could come up with a solution to that given time, but it's not a trivial issue and it's a significant roadblock to the idea.