Anthropic's Claude adds new memory features
Add Axios as your preferred source to
see more of our stories on Google.
Anthropic's Claude Sonnet 4. Image: Anthropic
Anthropic's Claude can now remember and incorporate previous conversations into all of your chats over weeks and months, as long as you're a paying subscriber.
Why it matters: Chatbots without memory are like brilliant interns who forget everything once the meeting ends.
Between the lines: Anthropic rolled out Memory to its Team and Enterprise plan users last month and is now rolling it out to Pro and Max plan subscribers.
- Users can see exactly what Claude remembers and can edit its memories.
- Claude competitor ChatGPT also has several different long-term memory features, even on its free tier.
Zoom in: Memory can make chatbots seem overfamiliar when users mix work and personal chats.
- Anthropic promises that if users are managing multiple Projects, Claude creates separate memories for separate projects. "Product roadmaps stay separate from creative writing," Anthropic says.
- The company says it ran extensive safety testing to confirm that memory wouldn't reinforce harmful patterns, lead to over-accommodation or attempt to bypass safeguards.
What they're saying: "We're building toward Claude understanding your complete work context and adapting automatically," Mike Krieger, Anthropic's chief product officer, said in a statement.
If you're a regular ChatGPT or Gemini user and you decide to work mainly in Claude, you can export your memories and import them into Anthropic.
Yes, but: Exporting data from a chatbot is about as easy as exporting it from Facebook or Google — which is to say, not easy.
What we're watching: Long-term memory could make chatbots genuinely helpful assistants, as long as the companies get privacy, context and control right.
