Service eligibility and regional restrictions
PuppyIP serves only compliant overseas businesses and their authorized personnel. Proxy services are not available in mainland China. The service may only be used for lawful business activities outside mainland China. Use of this service within mainland China is prohibited.
Hosting a proxy IP or server overseas does not change these restrictions. The service must not be provided to end users in mainland China through relaying, forwarding, sharing or resale. Before use, read the Terms of Service.
Key Takeaways
- Claude Chat and Cowork now share the same memory. Work context created in either interface can affect later conversations in the other.
- Memory is on by default for Free, Pro, and Max on web, desktop, and mobile; mobile apps need the latest version. Administrators control availability for Team and Enterprise, and individual users must still enable it themselves.
- Memory appears as Topics that can be read, edited, or deleted. Claude also updates topics during a conversation, rather than generating a summary only after the session ends.
- Sensitive topics are not saved by default. Users can opt to include some sensitive topics going forward, but this is not retroactive. Information such as government ID numbers, criminal history, and immigration status still will not be saved.
- Deleting a chat does not necessarily delete memory entries derived from it. When handling incorrect or sensitive information, check the original conversation, Topics memory, exported data, and team retention policy separately.
What changed: Cowork brings in context from another chat
On August 25, 2026, Anthropic announced that Claude Chat and Claude Cowork use the same memory. Cowork cloud tasks can use context formed in chat, and new information in Cowork may enter later chats. Memory is enabled by default for Free, Pro, and Max on web, desktop, and mobile; iOS and Android need the latest app version. For Team and Enterprise, administrators decide organizational availability, and individual users must still enable it after it becomes available. The announcement lists no regional rollout, so check version, plan, and administrator settings first if the entry point is missing.
A concrete example: you ask Cowork to draft a weekly report, and it uses an old project name or priorities from another job. The visible symptom is information in a new task that was not supplied in its current prompt. The cost can be incorrect documents, spread of sensitive context, and team review time. Deleting the current chat is not enough to clear the source. First open Settings > Memory, inspect Topics, and identify the entry before deciding whether to edit, delete, or pause it.
How the old and new experiences differ
Users may previously have understood chat memory as a feature of one interface or one session. Chat and Cowork now share it, and Claude updates Topics as conversations progress. This does not insert every complete chat verbatim into each task or give Cowork all organizational data. Available context remains limited by account, workspace, and product permissions.
Anthropic is gradually introducing the improved memory experience. Settings > Memory usually indicates the new experience; accounts still showing Memory under Settings > Capabilities may have the old one. Team and Enterprise migration timing also depends on administrator notices and organizational settings. Do not use a personal Free/Pro interface to draw conclusions for team members.
Steps to view, edit, and delete Topics
Open Settings > Memory and review the categorized entries and their details one by one. For outdated but useful content, use Tell Claude what to change or remove. For content that should no longer be used at all, choose Delete. Changes affect later conversations. Record the original value and reason for the change first; have the information owner review team terminology, customer names, and project status.
For temporary isolation, turn off memory or search past chats, then use a new conversation to check whether Claude still references the old context. Do not end the check merely by asking Claude “have you forgotten?” Return to settings to confirm the entry's state and check whether the original chat still exists.
Boundaries of the sensitive-topics setting
Sensitive topics such as health, race, ethnicity, religion, politics, and gender identity are excluded from memory by default. Users can opt into include sensitive topics in memory. This affects only future saves, does not add past content retroactively, and displays a notification whenever a related topic is saved.
Even with the setting enabled, Claude will not save certain categories, including sensitive identifying numbers such as Social Security or government ID numbers, criminal history, immigration status, and content that violates the Acceptable Use Policy. Do not treat the setting as a password vault, identity database, or compliance archive. Keep such data in approved systems.
Deleting chats, deleting memory, and exporting data are different actions
Anthropic's help documentation says that when a chat expires or is deleted, related memory entries created from it are not automatically deleted as well. Users can delete memory separately. Conversely, deleting a Topic does not delete its original chat. Check both locations when handling sensitive information instead of treating one delete button as a global erase.
All memory data is included in data exports, and Enterprise data-retention policies also apply to memory-related data. Import and export can support migration or backup, but import remains experimental and may not absorb everything completely. Remove unnecessary personal information before importing, and check each item afterward. A success message is not proof of completeness.
Checklist for individuals and team administrators
Individual users should check whether memory is enabled, whether Topics are accurate, whether the sensitive-topics setting matches their intent, and whether mobile apps are current. Run a nonsensitive check in both Cowork and Chat. Team administrators should also confirm the organization-wide capability setting, whether members can enable it themselves, retention policy, export and offboarding processes, and who handles incorrect memory.
If team members see different menus, first record the plan, client version, Settings path, and administrator policy. Do not start by changing proxies or repeatedly signing in. The announcement does not tie the new memory entry point to a particular IP. Investigate the network layer only when there is evidence of DNS, TLS, connection timeouts, or explicit regional support constraints.
Verify corrections, fall back, and know when to stop
After editing or deleting memory, ask a question in a new conversation without supplying the answer, then create a low-risk Cowork task and observe whether it still references the old information. Do not test with real identity documents, medical records, customer secrets, or production credentials. If old content persists, save the relevant chat reference and Topics state, then escalate to your organization's administrator or official support.
Stop if you cannot tell whether information came from memory or a current file, cannot locate a sensitive entry in settings, see a conflict between organizational policy and a personal setting, cannot verify an import, or find Cowork carrying unrelated project context into external deliverables. Disable memory or use manually supplied context that does not depend on it, and complete review by the data owner first.
Common misconceptions, related guides, and when to recheck
Common misconceptions are that Chat and Cowork have fully separate memories; deleting a chat automatically deletes memory; enabling sensitive topics imports old content retroactively; Team/Enterprise share personal-plan defaults; or a missing entry point proves a network block. Base your conclusion on the official announcement, current Settings path, and administrator policy.
If Claude cannot sign in or load at all, see IP checks before signing in and AI API error troubleshooting. Network products cannot change Claude's memory rules. Recheck both Chat and Cowork 24 hours after a change, and check the official help page a week later for completed updates to team migration guidance.
Sources
Frequently Asked Questions
Do Claude Chat and Cowork use the same memory?
Yes. Anthropic's August 25, 2026 announcement says Chat and Cowork now share one memory, and context formed on either side may affect later work.
Where can I view and delete Claude memory?
In the new experience, view, edit, or delete Topics under Settings > Memory. Accounts still showing Settings > Capabilities may have the old experience and should follow the corresponding help guidance.
Does deleting a chat also delete its memory?
There is no automatic guarantee. Official help says related memory entries may remain after a chat is deleted or expires. Delete the corresponding entries separately in Memory.
Will enabling sensitive topics add past chats to memory?
No. It affects only future saves and does not add past content retroactively. Certain sensitive identity information, criminal history, and immigration status are not saved even when enabled.
Is memory enabled by default for Team and Enterprise?
Administrators control organizational availability, and individual users still need to enable it. The improved experience is being introduced gradually, so team members may temporarily see different settings paths.
When should I pause Claude memory?
Pause memory and use explicit, manually supplied context when incorrect or sensitive content cannot be located, organizational policy is unclear, context crosses between unrelated projects, or import results cannot be checked.