Claude gains optional automatic memory, incognito chats

Anthropic is rolling out an automatic memory feature for its Claude chatbot that can recall details from past conversations without being prompted. Initially available to Team and Enterprise customers, the feature lets Claude surface people’s preferences, project context, and priorities to make responses more relevant across work use cases.

The company had previously added a manual memory prompt for paid users; the new update lets Claude proactively incorporate stored details and links those memories to “projects” (used by Pro and Teams) so the assistant can generate diagrams, website designs, graphics and other assets informed by uploaded files. Anthropic says the memory option is fully optional and that users can view, edit, or tell Claude what to focus on or ignore from their settings.

Alongside memories, Anthropic is launching incognito chats for all users; conversations in incognito won’t be saved or referenced later. The story notes that OpenAI and Google already offer cross-chat memories, and cites a New York Times report that linked ChatGPT’s memory rollout to more reports of “delusional” AI chats — a reminder of potential risks as chatbots keep longer-term context.

View Full Story