AI chatbot memories create algorithmic self-portraits of users
In brief
- AI chatbots like ChatGPT build personalized memories from user exchanges, including psychological data.
- A study reveals 96 % of memory updates are system-initiated, not user-driven.
- 28 % of stored memories contain personal data, 7 % of which are sensitive under GDPR.
Generative AI companies are integrating memory features into their chatbots to personalize responses, which researchers describe as creating « algorithmic self-portraits » of users. These systems collect personal data, including psychological information, classified as sensitive under GDPR.
A study by the Max Planck Institute shows that 96 % of memory updates in ChatGPT are initiated by the system, not the user. Among the 1,058 conversations analyzed in February 2026, 28 % of stored memories contain personal data, including 7 % in sensitive categories. Researchers highlight that these memories reflect not just facts, but also users' desires, intentions, and knowledge.
The research, presented in April 2026, exposes the opacity of data collection processes and privacy risks. OpenAI has since modified memory functionality, but researchers continue to warn about user control and data transparency issues.
What it changes for you
- Public SectorRegulators may tighten transparency and user control requirements for chatbot memories, affecting data protection policies.
This summary gives you the essentials. The full article, its figures and quotes are on Next.
Read the original articleAI-generated summary of the source, checked by rules. How we write · Report an error