Anthropic expands Claude's memory to Cowork and adds sensitive topic controls
Anthropic has updated Claude's memory feature so that it now carries context between the chatbot and its Cowork automation agent. Users can opt to include sensitive topics in memory, with notifications each time such data is saved. The company says Claude will not store certain personal identifiers like social security numbers.
The update unifies memory across Claude's chat interface and its Cowork automation agent, meaning context established in one environment now carries into the other. This follows a recent Chrome extension upgrade that brought Cowork into the browser side panel, making those conversations part of usage history. Memory remains enabled by default across Free, Pro, and Max tiers, with mobile users needing app updates to access the new controls.
Anthropic's sensitive topics toggle is opt-in, and the company explicitly excludes certain identifiers—social security numbers, criminal history, immigration status—from storage regardless of user preference. Users receive notifications whenever sensitive data is saved, and can edit or delete memories at any time. Past conversations are not retroactively added to memory.
This update could reduce friction for users who rely on Claude for recurring automated tasks, potentially increasing workplace reliance on AI agents. The sensitive topics control may ease privacy concerns, though users must actively understand what they are enabling. Broader adoption of persistent memory across tools could normalize AI retaining personal context, raising questions about data longevity and user oversight that may shape expectations for competing AI assistants.