I suspect this is just gibberish, but if not it poses an interesting confused deputy problem, what if AIs are accidentally trained on resources that weren't mea

I suspect this is just gibberish, but if not it poses an interesting confused deputy problem, what if AIs are accidentally trained on resources that weren't meant to be public? (E.g., imagine the open ai crawling bot had permission to the open ai private GitHub repos or something).