

The “independent, new and substantive statements” framing is the part I think travels furthest. Google’s defense was basically “we’re just summarizing,” but a summary that recombines facts into a new claim — this doctor was convicted — is authorship in any meaningful sense, so the old search-engine safe-harbor logic doesn’t really apply. If that reading holds across the EU it gets interesting fast for every company bolting an AI chatbot onto their search or docs. It’d be cheaper to cite sources verbatim with links than to defend a generated claim in court, which honestly would also make the products better. Also notable this is the second German court to land on the same side this year — one ruling is an outlier, two starts looking like a direction.
The support chat angle is the part that should worry people more than the AI prompts — customers write insanely private stuff to support because they assume there’s a human on the other end who reads it and forgets it. Instead it got dumped into the same 23TB bucket as everything else. The uncomfortable lesson is to treat any text you send to a third-party service as a permanent, potentially-public record. Doesn’t matter if it’s an AI prompt or a support ticket, it all just sits in someone’s storage account until somebody forgets to flip the auth flag. That’s not really an AI problem, it’s a ‘your data lives on someone else’s infrastructure’ problem.