Anthropic is bridging the gap between its AI workspaces with a major update that brings shared memory to Claude Cowork and Claude chat. Users no longer need to constantly re-explain context or manually prompt the AI to retain specific details when switching between tasks. The system will now dynamically remember information across both environments in real time.
Previously, Claude would only summarize conversations after they ended, requiring users to explicitly type commands to remember specific facts. Following this update, which Anthropic recently announced, the AI adds topics to its memory as the chat progresses. This creates a seamless workflow where a task executed in Cowork is instantly recognized in standard chat, and vice versa.
To maintain user control, all retained context is stored in a dedicated list of files under the Topics section within the Memory settings. From this dashboard, users can read, edit, or delete any saved memories. The system also allows users to pause or completely reset the memory function at any time.
While the memory feature is enabled by default, Anthropic has implemented strict privacy guardrails. Sensitive information - such as health data, religious beliefs, political views, or gender identity - is explicitly excluded from automatic saving. Users who want this data retained must manually toggle the 'include sensitive topics in memory' option.
However, certain data is permanently blocked from being saved, regardless of user settings. Identification numbers, criminal history, and immigration status will never be stored, adhering to the company's Acceptable Use Policy. The Claude Cowork memory update is currently rolling out to users on Free, Pro, and Max plans across web, desktop, and mobile platforms.
The Privacy-First Approach to AI Context
Anthropic’s implementation of shared memory highlights a growing divide in how AI companies handle persistent context. By explicitly blocking the automatic retention of sensitive data like health and political views, Claude is positioning itself as the safer alternative for enterprise and professional environments. This granular control is crucial for users who rely on Cowork for sensitive business tasks but want to avoid accidental data leakage into their general chat history.
Furthermore, the hard ban on storing identification numbers and criminal history demonstrates a proactive approach to compliance and security. As AI assistants become deeply integrated into daily workflows, the ability to seamlessly share context between a workspace and a chat interface - without compromising personal privacy - will likely become the industry standard that competitors must match.
