Anthropic unifies memory across Claude chat and Cowork, raising privacy concerns
Anthropic announced that Claude chat and Cowork now share a single memory system, allowing both to access user history. The feature is two-way and opt-out, raising privacy questions about what the AI retains. Claude Code is not included in this update.
The memory merger means months of accumulated chat context becomes immediately available to Cowork when it begins a task, and vice versa. Anthropic's examples show the system drawing on preexisting knowledge, such as a manager's preferred update format or conference details from earlier conversations. The company framed this as eliminating the need for rebriefing across repeated work.
Claude Code remains excluded from the unified memory system, with Anthropic declining to say whether it would eventually be included. The announcement reflects a broader industry pattern where both Anthropic and OpenAI are consolidating features into integrated desktop applications, moving beyond single-purpose tools toward more comprehensive AI workspaces.
This unified memory could reshape how users interact with AI assistants, making them more proactive and context-aware in daily work. However, the opt-out design means many users may unknowingly have their chat history accessed by Cowork, raising questions about data autonomy. Professionals handling sensitive information could face heightened exposure risks, while productivity gains may pressure users to accept the tradeoff. The exclusion of Claude Code suggests cautious rollout, but the trajectory points toward increasingly comprehensive AI retention of user activity.