“People can scrape your website, therefore you should just submit and allow your server to freely hand all your user’s posts over to meta upon request” is quite the take.

This. I’m sure it’s already happening. People training LLMs are already pointing their models towards ActivityPub.