[t-3]

When you say abused, I assume you mean either CSAM or copyrighted material?

Is there a hash database or something that could be queried to block known bad stuff? (would probably fail in the face of compression or encryption, but catching the low-hanging fruit would at least probably handle any potential legal liability?) Seems like something useful AI would actually work well for, if FBI/publishers/etc., would train a model and release it or host a service with an API.

[j16sdiz]

> catching the low-hanging fruit would at least probably handle any potential legal liability?

Actually No. It take 5 minutes to create a script to randomize a password, encrypt and upload.

Abusers have been doing the same for 20+ years and it is very effective.

[ttyprintk]

I think the question is not about the persistence of uploaders but the safe harbor of innocent hosters.

For example, hosting a password-protected binary with a cleared hash is one thing. The bad actor posting a password (any password) might be another.