tuckerm

This is no different from how it is already, though. If a user posts a CSAM link, your server immediately visits that link to scrape it, and you suddenly have that content on your server.

Of course, if you have a blocklist of forbidden sites, then your server will know not to scrape that site. And the same thing applies if we allow servers to send pre-scraped metadata. If the incoming message is for a blocked URL, just drop the message and don’t save any part of it.

The only new threat here is for a malicious server operator to modify outgoing messages to include CSAM metadata on a harmless link. However, that possibility already exists with any other content posted on the fediverse. If a malicious server is modifying its own outgoing messages, the only fix is to defederate from it.

Allowing servers to forward pre-scraped metadata to other servers does not introduce any new threats that do not already exist.

I’m a fan of the second one. My thought is, if you trust a server enough to accept messages from it, then you should also trust it enough to let it scrape the page title, article image, etc. for you, too.