The article was titled poorly and doesn’t have much detail in it (probably because the press release was vague itself) so I’m guessing. It sounds like Discord themselves will be scraping all the public servers with LLM’s in order to generate a “summary” of all the messages and put it on a public webpage for each server. If the big LLM companies can only get summaries then they can’t use those as training data.

It’s unclear from the article, as it says Discord is posting public-facing Server Web Pages that “contain summaries”. No info on how the summaries are generated.

I probably guessed wrong before, as if it’s not access to the raw messages then there’s not good AI training data. Is Discord going to be scraping the servers and summarizing with LLM’s? That sounds likely.

Turn your server private, invite-only

deleted by creator

This absolutely means Google is going to use the messages to train Gemini.

edit: rather, whatever LLM Discord will be using to “summarize messages” will be using them as training data. Not sure which one they’re using.