Claude’s prompt turns copyright refusals into conversation memory
Anthropic’s published Claude prompts now make copyright enforcement unusually explicit: no song lyrics, poems, book passages, recognizable characters, or logos, even when a user asks piecemeal or tries a reworded workaround. The refusal persists for the rest of that conversation, while Claude offers analysis instead. That is a meaningful shift from a single blocked answer to a session-level policy, arriving as music publishers sue Anthropic over lyric-training data.