What ChatGPT Thinks It Knows About You Is Affecting Its Answers. Here’s How to Change That

ChatGPT’s default memory now automatically infers and retains information about you, influencing future answers.

ChatGPT has updated its memory feature, previously requiring explicit user commands to retain information. The latest iteration, introduced in June and referred to as 'dreaming,' allows the AI to automatically build, edit, and sift through memories without direct user input. This means ChatGPT now refines its understanding of users over time, aiming to provide more personalized and contextual responses. However, this automation means the AI, rather than the user, largely determines what is remembered, potentially leading to inaccurate assumptions. Users can now access and manage these memories through the 'Personalization' settings in the web and mobile apps. Options include disabling the memory feature entirely, editing specific memory summaries by correcting or marking information for non-mention, or completely deleting all stored memories. Temporary chats also exist, where no conversation data is saved to memory.

The shift from explicit user commands to automatic memory management in ChatGPT represents a significant step towards more autonomous AI agents. By 'dreaming,' the AI actively curates user profiles, aiming to enhance personalization but sacrificing user control over what context is retained. This move mirrors trends in other AI systems where implicit data collection and inference are becoming standard for tailoring user experiences.

Users now face a trade-off between convenience and privacy, as ChatGPT's memory feature makes assumptions about their preferences and knowledge. The ability to manually edit or delete these inferred memories offers a degree of recourse, but the default setting prioritizes AI-driven contextualization over explicit user consent for every piece of retained information.