Oddbean new post about login | logout Does anyone know why isn't there some public database of fuzzyhashes or neuralhashes or something we can all rely on to purge CSAM material from relays and whatnot? Is there a law against that or what?
@fiatjaf It's gatekept so only big companies are immune from those kinds of lawsuits. Section 230 doesn't cover crime, after all. It's an easy way to take someone down.
There are, but they have their own problems. They're a black box, by design. (How did this particular fuzzy hash end up here? You're not allowed to know.) Also companies that host these lobby governments to mandate everyone use their services, so they are constantly trying to rent seek. One of them, Thorn, is a sponsor of chat control for instance. https://fortune.com/europe/2023/09/26/thorn-ashton-kutcher-ylva-johansson-csam-csa-regulation-european-commission-encryption-privacy-surveillance/
IWF: https://www.iwf.org.uk/our-technology/our-services/image-hash-list/ This is an interesting problem because we must first obtain CSAM in order to purposefully train a CSAM detection model, so the first step is illegal territory, hence the strict license for this particular academic research... This makes it hard to open source, unlike any previous only NSFW detection. See also: https://nsfwjs.com/