Anthropic has released an enhanced memory feature set for its AI chatbot, Claude, which is rolling out to all paying subscribers. The update is an expansion of a capability first introduced in August that enabled Claude to remember past conversations for recurring topics.
The upgraded memory system is designed to learn a user’s individual preferences and work patterns over time. Anthropic states this will translate to the chatbot developing a better understanding of how a person works. The company describes its implementation as “fully transparent,” noting that users will see an “actual synthesis” of what Claude has recorded, as opposed to “vague summaries” found on other platforms. Users can also edit the chatbot’s memory directly through conversational prompts. This functionality builds on the initial August update, which allowed users to reference specific chats to avoid repeating information when revisiting a subject.
A new feature called Projects has been introduced to compartmentalize the data Claude collects. When users group conversations together using this function, the chatbot establishes a distinct memory space for each grouping. This structure ensures that information saved from professional work conversations does not carry over into personal chats.
Anthropic also incorporated interoperability, making it possible for users to import saved memories from ChatGPT or Gemini into Claude. The update also enhances tool integration, allowing smoother use of features such as Math solver for contextual problem-solving within projects. In addition, contextual information that Claude saves can be exported for use on other AI platforms.
Ahead of the announcement, Anthropic conducted extensive testing to assess if the new memory capabilities could lead to an increase in sycophancy or the generation of more harmful content. “Though this testing, we identified areas where Claude’s responses needed refinement and made targeted adjustments to how memory functions,” the company said in a statement. “These iterations helped us build and improve the memory feature in a way that allows Claude to provide helpful and safe responses to users.”
The new memory capabilities are available immediately for all Max subscribers, and access for Pro users will follow in the coming days. The feature is entirely optional and is not enabled by default. To use it, individuals must manually activate the function by using a toggle within the chatbot’s settings menu.
