How the shared memory functions in practice
Anthropic has activated shared memory between its Claude AI assistant and Cowork productivity tool, meaning information gathered in one service is now accessible in the other. The feature went live on August 25, 2026, with no user option to disable or compartmentalize the data flow between the two platforms. Claude Code remains isolated from this sharing arrangement for the time being.
Breaking news
Echo Software Acquires Minimus Assets to Strengthen AI‑Driven Container Security
Plaud One Redefines Headphones for AI Note‑Taking
Apple Event Logo Suggests iPhone 18 Pro May Feature New Colors and Camera Upgrade
Plaud unveils smart earbuds that capture audio and execute tasks automaticallyThe integration allows Claude to recall interactions from Cowork and vice versa, creating a continuous context across tasks like document drafting, meeting summaries, and code development. Anthropic confirmed the change in response to inquiries, stating that the shared memory is designed to improve workflow efficiency by reducing the need for users to repeat information. However, the lack of user controls has raised questions about data privacy and consent, particularly for individuals using the tools in professional or sensitive environments.
Can users trust this level of data integration?
When a user discusses a project in Cowork, Claude can later reference those details during a coding session or general query, even if the conversation occurred hours or days prior. Similarly, insights from Claude’s interactions—such as preferences expressed during brainstorming—can inform Cowork’s suggestions for task prioritization or document formatting. Anthropic emphasized that the memory is not used for training its models but is stored temporarily to enhance personalization within the user’s session. The company said it is reviewing feedback but has not announced plans to add opt-out mechanisms or data separation controls.
Privacy advocates have noted that the absence of granular controls could pose risks for users handling confidential information, especially in regulated industries like healthcare or finance. While Anthropic maintains that data remains encrypted and is not shared externally, the inability to segment memory between services means users must either accept full integration or avoid using both tools concurrently. The company has not disclosed how long shared memories are retained or whether they are linked to identifiable user profiles beyond session-based context.
Is there a way to turn off shared memory between Claude and Cowork? No, Anthropic confirmed there is currently no option to disable or separate the shared memory between Claude and Cowork. Users who wish to prevent data flow between the services must use them independently or avoid linking their accounts.
Frequently Asked Questions
Does this shared memory apply to Claude Code as well? No, Claude Code remains excluded from the shared memory system at this time. Anthropic stated that the coding-focused variant operates with isolated context and does not exchange data with Claude or Cowork under the current setup.
How does Anthropic ensure the shared memory is not misused? Anthropic says the shared memory is used solely to improve real-time user experience and is not incorporated into model training. The data is encrypted during storage and transmission, and access is limited to the user’s own sessions across the linked services.
