Claude Now Remembers You in One Place Instead of Two
Anthropic merged the memory systems behind Claude chat and Claude Cowork on August 25. It is on by default, sensitive topics are off by default, and you can read and delete everything it keeps.
Anthropic announced on August 25 that Claude’s memory now works across both of its surfaces at once. Until now, anything you told Claude in a normal chat stayed in chat, and Claude Cowork, the version that works with files and tasks on your computer, started from scratch every time. From this week they share one memory, so context you gave in one place shows up in the other.
The rollout covers Free, Pro and Max plans on web, desktop and mobile, and it is switched on by default. Enterprise administrators can turn it off for their organisation. Claude now also writes to memory continuously during a conversation rather than only at the end of a session, and it organises what it keeps by topic, which you can browse, edit or delete item by item. On the privacy side, Anthropic says Claude will not store sensitive categories such as health information, race, political views or gender identity unless you explicitly switch on an “include sensitive topics in memory” option, and that some categories are blocked outright and cannot be stored at all: government ID numbers, social security numbers, criminal history and immigration status.
What’s actually going on here: memory is the least glamorous and most useful feature in an AI assistant. Without it, every conversation starts by re-explaining who you are, what project you are on and how you like things written, which is exactly the friction that makes people give up on these tools. Having two separate memories was worse than having none, because you could never predict which version of Claude knew what. The interesting part is not the merge itself but the shape of the controls around it: a topic-by-topic list you can actually read, an opt-in rather than opt-out for sensitive data, and a hard block list. That is a response to a real and reasonable worry, namely that a system quietly accumulating notes about you is one leak or one policy change away from being a problem.
What this means for you: if you use Claude for anything ongoing, this should make it noticeably less repetitive, and it is worth opening the memory view once to see what it has actually written down about you. It reads more bluntly than most people expect. If you would rather it kept nothing, the setting to turn memory off is there, and on a company account your admin controls it. For everyone else, treat this as the current benchmark for what to ask of any assistant that offers memory: can you see the whole list, can you delete a single entry, and is sensitive data opt-in rather than opt-out? Those three questions separate a memory feature you can trust from one you cannot.
Sources
Source: https://techcrunch.com/2026/08/25/claude-cowork-finally-remembers-what-you-told-the-app-in-chat/
One Bad Web Page Could Quietly Rewrite Your Local AI Model
Researchers at Oasis Security found that Nvidia's NemoClaw left a local Ollama server open to the whole network. Visiting a single malicious page was enough to plant hidden instructions in the model.
