Anthropic has also changed how memory is updated. Users no longer need to explicitly tell Claude to remember something for it to become part of stored context. Instead, memory can be updated as conversations continue, allowing Claude to retain information it determines may be useful later. Users can pause or reset memory through the Memory section in settings. Anthropic is also making saved information easier to inspect by organizing memories into topics that can be viewed, edited or deleted individually.
Another addition gives users the option to include sensitive topics that Claude would normally avoid storing. That setting is disabled by default and must be turned on manually. Anthropic said the feature is intended for cases where information that may be considered sensitive could still be useful in future conversations. One example given by the company is remembering a gluten allergy when helping with meal planning. “By default, Claude does not store topics related to personal or sensitive subject matter, like your health, race, ethnicity, religious beliefs, politics, gender identity, and other similar areas,” Anthropic said.
When sensitive-topic memory is enabled, Claude will notify users when information in that category is saved. Anthropic said the feature does not retroactively add sensitive information from earlier conversations, and users can disable it or remove saved memories at any time.
Some categories remain excluded even when the setting is active. Claude will not store certain information including government identification numbers, criminal history, financial account numbers or immigration status. The assistant will tell users when information cannot be added to memory.
The broader memory system is enabled by default for Free, Pro and Max accounts across web, desktop and mobile. Users on iOS and Android need the latest version of the Claude app to access the new features.
The update moves Claude closer to a persistent cross-product memory system rather than one limited to individual conversations. For Cowork in particular, the change reduces the need to rebuild context each time users switch from chatting with Claude to assigning the agent a task.
This analysis is based on reporting from CNET.
Image courtesy of Anthropic.
This article was generated with AI assistance and reviewed for accuracy and quality.
