Anthropic has launched a new memory feature for its Claude chatbot, allowing users to retrieve and summarize previous conversations on demand.

The feature is available for Claude’s Max, Team, and Enterprise subscribers across web, desktop, and mobile platforms, with plans to expand to other tiers.

Users can activate the memory function in the settings under “search and reference chats,” and it keeps different projects and workspaces separate.

Unlike OpenAI’s ChatGPT, Claude only accesses past chats when requested and does not create a user profile, according to Anthropic.

Anthropic and OpenAI are competing to launch new features and grow their user base amid rising investment and debate over chatbot memory.

🔗 Source: The Verge

🧠 Food for thought

1️⃣ Memory features become the new battleground for AI user retention

The launch of Claude’s memory function represents an escalation in the competitive dynamics between leading AI companies, where memory capabilities are increasingly seen as crucial for creating user “stickiness.”

Both Anthropic and OpenAI have been racing to roll out competing features like voice modes and larger context windows, with OpenAI launching GPT-5 just last week while Anthropic seeks funding at a $170 billion valuation.

Memory features specifically help improve user satisfaction by recalling preferences and past inquiries, leading to more relevant responses without users needing to repeat information across sessions1.

This competitive focus on memory reflects a broader industry understanding that personalized experiences drive user engagement and platform loyalty in the rapidly commoditizing chatbot market.

2️⃣ Reactive versus persistent memory reveals different AI safety philosophies

Claude’s approach to memory, only retrieving past conversations when explicitly asked, represents a fundamentally different design philosophy compared to ChatGPT’s persistent memory that continuously builds user profiles.

This reactive approach appears to prioritize user privacy and control, as users must actively request memory retrieval rather than having their conversations automatically analyzed and synthesized.

The distinction reflects broader concerns about AI memory capabilities, as enhanced memory features create a balance between personalization benefits and privacy risks that users must navigate2.