OpenAI is enhancing the personalization of its ChatGPT by introducing a “memory” feature. This innovation will enable ChatGPT to retain information about users and their interactions over time.

There are two methods for this memory function to operate. Users have the option to directly instruct ChatGPT to remember specific details about them: preferences for coding in JavaScript, the name of their boss, or their child’s allergy to sweet potatoes.

Alternatively, ChatGPT can gradually accumulate this information through ongoing conversations, learning about users as they pose questions and receive responses. The objective is to make interactions with ChatGPT feel more customized and intelligent, eliminating the need for repetitive reminders.

This ability will also extend to custom GPTs on the GPT Store with each chatbot possessing unique abilities to remember information about the user. For instance, Books GPT will remember user preferences on books for better recommendations and similarly, Tutor Me could tailor a more effective long-term educational plan based on user information. The potential applications for memory within the GPT Store are vast.

This raises privacy concerns as well since ChatGPT will start learning everything about users and create a detailed profile based on their personal data. Fortunately, the chatbot will also allow users to remove memories whenever needed. There will also be an option to completely opt out of the memory option. OpenAI also says that the system has been trained to not remember sensitive information such as health.

Users can ask the chatbot what it remembers about them and tell the AI to forget about it or do it manually from the Manage Memory section of settings. Additionally, OpenAI is introducing the “Temporary Chat” feature, designed as a privacy mode for users wishing to engage in conversations that do not contribute to the accumulated memory ChatGPT has of them.

But keep in mind that ChatGPT will keep its memory feature on by default and the company says this data will be used to train its models going forward. ChatGPT Enterprise and Teams will be able to opt out of this data collection as always.

OpenAI says that this feature is only available to a small portion of users at the moment, but will roll out to a wider audience soon.