Anthropic is rolling out an update to Claude’s memory feature that works across both individual conversations and dedicated work sessions. The change aims to make the AI assistant more consistent over time. While the model has been able to recall details from previous chats since last summer, that functionality has often been fragmented. Today’s update bridges the gap, allowing the assistant to access stored information whether a user is in a standard chat or using the Claude Cowork computer use agent.
Connecting Chat and Computer Use
The update ensures that when Cowork runs a task in the cloud, the memories it pulls from chats remain available. This connection fixes a long-standing issue where the AI had to restart tasks from scratch. The timing aligns with Anthropic’s recent move to make Cowork accessible through its Chrome browser extension. Conversations held in the Chrome side panel now become part of the broader usage history.
Privacy controls have also been improved. Users can now review memories organized by topic and edit specific details directly from the settings menu. A new toggle labeled “include sensitive topics in memory” lets users decide whether Claude should store information related to personal or sensitive subjects. By default, the assistant avoids saving details about health, race, ethnicity, religious beliefs, politics, or gender identity.
Enabling the toggle changes how the model behaves, though Anthropic has set clear limits. Claude will not add past conversations to its memory retroactively, and users can disable the feature or delete entries at any time. The system also blocks certain data types, such as social security numbers, criminal history, and immigration status, regardless of user settings.
The expanded memory function offers a practical advantage. It does not store every detail but instead filters out irrelevant information. This approach prevents the model from being overwhelmed by trivial data while preserving context needed for complex or ongoing projects. The feature is enabled by default for Free, Pro, and Max accounts, though mobile users must update their apps to access it.
Selective retention helps maintain efficiency.
Anthropic’s approach reflects a balance between utility and user control. The assistant’s ability to remember key details across sessions could streamline workflows, especially for those managing multiple projects. For example, users working on a long-term research task or coding project may find the AI’s improved recall reduces repetitive explanations. The update also aligns with growing expectations for AI tools to handle context more naturally, similar to how newer hardware standards improve data transfer consistency.
While the feature is designed to enhance productivity, Anthropic has emphasized transparency. Users can audit stored memories and adjust settings to match their comfort level. The company’s decision to exclude sensitive topics by default addresses concerns about privacy without requiring manual oversight. For those who need the AI to remember more, the toggle provides flexibility without compromising security.