Anthropic has merged the memory systems behind its Claude chatbot and Claude Cowork, its agentic work tool, so that what you tell one is available to the other.
According to ZDNET's David Gewirtz, whose reporting was aggregated by Techmeme, Claude chat and Cowork now run on a single memory system, and your chat history becomes available to Cowork unless you opt out. Memory updates as you chat, and there are controls for sensitive data.
The Verge reports that Anthropic added memory to Cowork using the same system the chatbot already uses, making it easier to pick up tasks without re-explaining important information — and that memory is on by default. Engadget frames the same update from the control side, reporting that Anthropic is giving users more say over what memories Claude saves.
SiliconANGLE reports that the changes let users see everything Claude remembers "topic by topic" and edit it, with the update also aimed at protecting sensitive topics. The New Stack summarized the move simply as Anthropic giving chat and Cowork one memory.
The tension in the coverage is worth noting. The convenience case is straightforward: an AI that already knows your projects, your preferences, and your context doesn't make you start from scratch every session. But ZDNET's framing raises the privacy question directly — what an AI remembers about you, and whether the tradeoff is worth it, especially when the sharing happens by default rather than by request.
That default is the crux. Casual chat and work sessions are different contexts, and merging them means information volunteered in one setting can surface in the other unless a user actively intervenes.
This matters because persistent, cross-product AI memory is quietly becoming the norm, and the burden of drawing the line between personal and professional context is landing on users who have to know the opt-out exists.