AI Memory Sharing: Anthropic’s Claude And Cowork Will Remember You—Or Not

📊 Full opportunity report: AI Memory Sharing: Anthropic’s Claude And Cowork Will Remember You—Or Not on ThorstenMeyerAI.com — validation score, market gap, and execution plan.

Anthropic has changed its default privacy setting to enable shared memory to enable shared memory between its AI products, Claude and Cowork. This means that information the AI remembers about a user will now automatically carry over between the chat assistant and the work automation tool unless the user explicitly opts out. The move increases convenience for users but raises privacy concerns for those wary of cross-product data flow.

Previously, memory was scoped more narrowly, requiring users to manually choose to share context between Claude and Cowork. Now, sharing is the default, with an opt-out option available in Claude’s settings, according to Anthropic. This change impacts anyone using both products with memory enabled, as personal details, preferences, and ongoing project information will be accessible across both tools unless users disable the feature.

Anthropic describes shared memory as a way to reduce repetitive input and improve assistance quality, especially for multi-step tasks. Learn more about how AI memory sharing impacts privacy in this detailed report. For example, a user who shares their workflow with Claude will find Cowork better equipped to carry out related actions without re-explanation. However, several details remain unclear, including which specific memory types are transferred, whether users can selectively share certain memories, and how existing memories are handled during the transition.

Implications of Default Memory Sharing for Privacy and Convenience

This move signifies a shift in privacy defaults, as most users tend to accept default settings, making cross-product memory sharing the norm. For users prioritizing privacy, this means personal or sensitive information shared in one context could now be accessible to another product without explicit consent. Conversely, users seeking seamless workflows benefit from reduced repetition and more integrated assistance. The decision also aligns Anthropic with industry trends, as competitors like OpenAI and Google expand persistent memory features, raising ongoing questions about data use, retention, and security.

Background on AI Memory and Anthropic’s Product Evolution

Anthropic introduced memory features for Claude to enable the AI to retain information across conversations, such as user preferences, recurring projects, or writing styles. The feature was designed to be inspectable, allowing users to review and delete stored memories. Cowork, Anthropic’s agentic AI tool, is designed to execute multi-step work tasks, making shared memory a logical technical development to improve performance and user experience.

Prior to this change, users had to actively enable memory sharing, which limited cross-product data flow and provided more control over privacy. The new default shifts this control, making memory sharing automatic unless users intervene. The change arrives amid broader industry debates about persistent AI memory, data security, and user privacy, with rivals also expanding their memory capabilities.

“Users can opt out of memory sharing between Claude and Cowork in settings.”

— Anthropic spokesperson

Unresolved Details About Memory Transfer and User Control

Several key questions remain unanswered. It is not yet clear which specific types of memories — personal, professional, incidental — are shared automatically. It is also unknown whether users can selectively share or exclude certain memories, or if the control is a single toggle. Additionally, how existing memories are migrated or deleted across products, and whether shared memories are used for model training, remain unaddressed. The timing of full rollout and whether administrators can control this setting on managed accounts are also still developing.

Next Steps for Users and Industry Response

Users should review their Claude and Cowork settings this week to disable memory sharing if desired. Anthropic is expected to provide further clarification on selective sharing, memory management, and privacy policies. Industry observers will watch how competitors respond to Anthropic’s default change, especially regarding privacy protections and user control, as shared memory becomes a competitive factor among AI providers. Future updates may include more granular controls or transparency features.

Key Questions

What specific information is shared between Claude and Cowork?

It is not yet confirmed exactly which memories are shared, but it includes user preferences, ongoing project details, and contextual information stored by Claude. The scope of sensitive or incidental data transfer remains unclear.

How can I prevent my memories from being shared?

Users should go to Claude’s settings and disable the memory sharing option. It is recommended to review existing memories and delete any information they do not want to carry over to Cowork.

Does shared memory affect data used for training?

Anthropic has not addressed whether shared memories are used for model training or how deletion in one product impacts the other. This remains an open question.

Will this change be rolled out to all users immediately?

The rollout appears staggered, so some users may not see the new default or notification right away. Further updates from Anthropic are expected.

Can administrators control this setting for managed accounts?

It is not yet clear whether account administrators can enforce or restrict memory sharing defaults. Users should verify account controls if applicable.

Source: ThorstenMeyerAI.com