The memorization mechanism itself has also changed: previously Claude formed memory from a dialogue extract after its completion, now it adds topics to memory directly during the conversation, even before the chat is closed. This speeds up the exchange of context between chat and Cowork in real time.
What is stored and how
class=»notranslate»>__GTAG5__ Anthropic also gave users access to the contents of the memory — you can see what exactly the AI remembered for each topic, and, if desired, correct or erase the record.
By default, Claude does not store sensitive data — information about health, race, ethnic origin, religious views, political preferences, gender identity. You can enable their collection manually through the “consider sensitive topics in memory” switch, and the application will notify the user each time such saving is done.
A separate category of data is not saved at all under any settings: government document and social security numbers, criminal records, immigration status and everything that falls under the acceptable use policy of the service.
The function is already enabled by default on free, Pro and Max plans — on the web, desktop and mobile applications. To do this, iOS and Android owners need to update the application to the latest version.
Sources:
techcrunch.com
anthropic.com
claude.com
Xrust Claude was taught not to forget what you told him
- Если Вам понравилась статья, рекомендуем почитать
- TickTick: как пользоваться планировщиком задач, привычками и Pomodoro
- DaisyDisk for Mac: How to Find Large Files and Free Up Space Safely








