Claude organizes its retained information around specific topics, and users can manage these memories directly through the chatbot’s interface. According to Anthropic, the simplest method for editing is via the web client at claude.ai.
To adjust memories on a computer, navigate to Settings from the profile menu, select Memory, and locate the Topics section. From there, users can click the Edit icon to modify a specific memory using the prompt bar, or use the Delete icon to remove a topic entirely.
Mobile users can access these controls by tapping the profile icon, selecting Capabilities under the App section, and then tapping Memory files. Once a topic is selected, users may describe changes in the prompt bar or tap the trash icon to delete the entry.
Following a recent update, Anthropic introduced an optional setting called “Include sensitive topics in memory.” By default, Claude excludes memories related to health, race, ethnicity, religion, politics, or gender identity. Enabling this option allows the AI to record such sensitive data, though users will see a notice whenever a sensitive memory is generated. Anthropic noted that older versions and the mobile app do not display this notice and will not save sensitive topics.
Certain information remains protected regardless of settings; for instance, government ID numbers are never saved, and the system does not retroactively process past conversations. Users who wish to opt out of memory generation can toggle off “Generate memory from chats.” This provides two choices: pausing memory features or resetting them completely. Anthropic warns that resetting is irreversible, meaning re-enabling the feature will start Claude’s memory from scratch without any prior context.
Finally, granular control over what the AI remembers. That sensitive topics toggle is a game changer for privacy.
Wait, does this mean I need to manually review every memory? I just want it to forget everything and start fresh!