Consider an AI chatbot platform. If this platform doesn’t allow users to create characters (that are stored on their centralized service) that infringe IP, but allows users to import character definition files that do so from external sites (that are created by community members, and may contain copyright-protected characters), without permanently storing (or storing for private use), is the owner responsible for IP infringement?
It’d be interesting to see how something like Hyphanet (nee Freenet) interacts with such a requirement, as data is stored in a redundant distributed hash table that spans the whole network, there’s no way to contact the operator of a given node, redundant storage of data, and no real mechanisms for taking any content off the DHT once uploaded.