Claude's Memory Is Now Shared Between Chat and Cowork: What It Means for Your Business
Claude now uses a single memory across both chat and Claude Cowork, as of August 25, 2026. This means that when you discuss work in chat and then hand it to Cowork to continue, the system starts from the existing context right away with no need to explain it again. There is also a settings page where you can read, edit, and delete everything Claude has remembered.
Previously, Claude in chat and Claude Cowork kept separate memories, so people doing real work kept running into the same problem: you would plan a project in detail in chat, and when it was time to tell Cowork to start doing the work, you had to explain everything from the beginning again. This update removes that step and adds controls users can actually act on: you can see everything the system remembers, correct it, and delete it at any time, as Anthropic states in its Memory update announcement (Anthropic).
How does shared memory between chat and Cowork work?
There is one memory, and it flows both ways. What you discuss in chat is available when you hand work to Cowork, and what happens while Cowork is working can flow back into chat as well. Whether you use the web, desktop, or mobile, they all draw on the same memory.
In practice, what changes is where work starts. Previously, every new session began from zero. Now it begins from what the system already knows, such as who this client is, what report format the team prefers, or which step a project is stuck on. For teams that work in cycles, such as content or consulting work, the time saved comes from not having to re-brief rather than from faster processing.
Claude also adds to its memory on its own during a conversation when new details come up, and users can give a direct instruction such as "Remember that..." to save a specific item.
Where can you view, edit, or delete what Claude remembers?
Go to Settings > Memory. Everything Claude remembers is stored as separate files by topic in the Topics section. You can open each one to read what it says, edit it to correct it, and delete it one topic at a time. It is not hidden data that you cannot inspect.
What makes this useful in practice is that these files are short, and one edit takes effect everywhere. If the company changes its name, someone changes job title, or an old client's details are no longer correct, you fix it in one file and every later conversation uses the correct information. You do not have to correct it in the middle of each conversation.
We recommend opening this page and reading through it once before you decide anything else, because seeing your own context laid out as text is the fastest way to judge whether you want the system to remember more or less.
What settings should you use if you do not want Claude to remember personal matters?
By default, Claude does not save topics considered sensitive, such as health, ethnicity, religious beliefs, politics, or sexual identity. If you want it to remember them, you have to turn on the "Include sensitive topics in memory" switch yourself in Settings > Memory.
Turning this switch on only applies going forward. Topics discussed before you turned it on are not saved retroactively, and a notification is shown every time the system saves something in this group. If you turn the switch back off, items saved in this group are deleted. On mobile, you must update the app to the latest version first. Otherwise you will not see the notification and the system will not save anything (Anthropic Help Center).
Some categories are never saved even with the switch on: personal identification numbers, criminal history, and immigration status. This matters for organizations that handle employee or customer data, because it is a line the system draws itself that users cannot turn off or on.
Should a team or organization turn on Memory?
For Team and Enterprise accounts, Memory and sensitive topics are two separate switches, and both are off by default. An Owner or Primary Owner can turn on Memory for the whole organization at Organization Settings > Capabilities, but turning it on at the organization level does not turn on sensitive topics with it.
This structure means an organization can set the boundaries but cannot decide on employees' behalf about their personal information. Even when an Owner allows sensitive topics at the organization level, nothing is saved until each employee turns on their own switch, and an Owner cannot read an employee's individual memory.
Our recommendation for organizations considering this is to turn on Memory first, because the benefit is clear and the risk is low. Then communicate with the team about sensitive topics separately, letting each person decide for themselves, and tell them how to view and delete their own data at Settings > Memory. Letting employees know they can check at any time reduces hesitation more than explaining how safe the system is.
From our experience teaching Claude and Claude Cowork to organizational teams in northern Thailand, including at Claude Connect Chiang Mai in July 2026, the question executives ask most often is not what AI can do, but where the company's data goes and who can see it. This update answers that question more directly than before, because the answer is now a screen you can actually open and look at.
Frequently Asked Questions
Is Memory already turned on, or do I have to turn it on myself? For Free, Pro, and Max accounts, it is turned on by default. You can check at Settings > Memory. For Team and Enterprise accounts, it is off by default, and an Owner must turn it on at Organization Settings > Capabilities before employees can manage their own settings.
Why can't I see the "Include sensitive topics in memory" switch? There are 3 common causes.
- The main Memory setting is not turned on yet. This switch sits under Memory.
- You are using an organization account where the Owner has not yet enabled the permission.
- You are using an older version of the mobile app. Try checking against the claude.ai website first.
If I delete a memory topic, will Claude remember it again? Deleting applies to information that has already been saved, but if you discuss the same topic again in the future, the system may save it again as normal. If you do not want a topic remembered at all in the long term, tell Claude directly not to remember it, or turn off memory creation from conversations in Settings.
If I turn on sensitive topics, will it go back and pull information from old chats? No. The system saves going forward only. Topics discussed before you turned on the switch are not saved retroactively. If you want the system to know about them, you need to tell it again after turning on the switch, or type an instruction to remember that topic specifically.
Can a company Owner see employees' memories? No. An Owner can only control the permission level, meaning whether to turn the feature on or off for the whole organization, but cannot read the list of an individual's memories. Each employee manages their own memory content, including viewing, editing, and deleting it.
Can I move my memory from another AI to Claude? Yes. Claude has a memory import and export feature for moving context from other AI providers or backing up your data for later use. This feature is still under development, so we recommend checking the results every time after an import to confirm the information is correct and complete.
Author: Tanyalak Techapeompol Managing Director, Custocare, an AI and Digital Transformation consultancy with 17 years of IT experience Former CIO of Oasis Spa Group, overseeing IT systems across 14 branches in 4 provinces · AI speaker and guest lecturer for Chiang Mai University and private-sector organizations · Holder of 6 Anthropic certifications, including Introduction to Claude Cowork (Verify certifications)
Published: 27/08/2026 | Last updated: 27/08/2026
References