Claude adds memory feature for paid subscribers

Anthropic is rolling out a “memory” upgrade for Claude that enables the AI chatbot to remember past conversations without prompting, starting with Max subscribers today and expanding to Pro users over the coming days. This feature brings Claude closer to parity with rivals ChatGPT and Gemini, both of which introduced memory capabilities last year, as AI companies compete intensely to retain users through persistent, personalized experiences.

What you should know: The memory feature offers complete transparency, allowing users to see exactly what Claude remembers rather than receiving vague summaries.

• Users can toggle specific memories on and off or edit them through natural conversation, such as telling Claude to focus on certain memories or “forget an old job entirely.”
• The feature includes “distinct memory spaces” to keep different conversations and projects separate, preventing memories from bleeding across personal and professional uses.
• Anthropic aims to reduce switching friction by allowing users to import memories from ChatGPT or Gemini through copy-and-paste, with no lock-in policies for exporting memories.

The big picture: Claude has lagged behind competitors in memory functionality, only gaining the ability to remember past conversations in August 2024, and even then requiring explicit user requests to remember information.

• Team and Enterprise users have had access to the memory feature since September, while the rollout to individual paid subscribers represents a broader push for user retention.
• All major AI chatbots are now competing on memory features as a way to encourage users to stick with one platform rather than starting over with another bot.

Why this matters: Memory functions are becoming a critical differentiator in the AI chatbot market, but they also raise concerns among experts about potential mental health implications.

• Some specialists warn that chatbot memory could help sustain or amplify delusional thinking and other mental health concerns, colloquially referred to as “AI psychosis.”
• The concerns are particularly relevant given the sycophantic tendencies of some AI models, which could potentially reinforce problematic thinking patterns through persistent memory.

What’s next: Anthropic has not indicated whether the memory feature will eventually become available to free Claude users, keeping the premium functionality limited to paid subscribers for now.