Anthropic updates Claude’s memory to enhance customization and protect sensitive topics

Anthropic, an artificial intelligence startup, has announced updates to its Claude AI product that enhance customization and protect sensitive topics. Users can now view everything Claude remembers, organized "topic by topic," and edit or delete any of that information. By default, Claude does not store sensitive subjects such as health concerns, race, gender identity, religion, or political beliefs.

However, users can enable a toggle to add specific topics to Claude's memory. The AI's memory will now persist across both chat and Cowork, Anthropic's desktop AI agent capable of performing multistep tasks. With this update, Claude's memory remains consistent across multiple platforms and contexts, so a user does not need to repeat information each time they interact with the AI.

For example, if a user builds a report for their manager using Cowork, Claude can recall all relevant context from previous discussions about their manager's preferences and desired communication style. Users can easily adjust these preferences in Claude's memory or revert to previous settings if desired. Anthropic also noted that while Claude typically ignores sensitive topics automatically, users will be alerted when sensitive topics are included in the memory.

The AI will not save information that is highly private, such as Social Security numbers or criminal history, and will not store topics that violate Anthropic's Acceptable Use Policy, which includes safety, emotional harm, privacy, and identity guidelines.