OpenAI has changed the permission setting method for new health plugin connections in ChatGPT. Moving forward, when connecting a new health plugin, ChatGPT will follow the user-specific global plugin permission settings by default. If not changed separately, the default permission will be set to 'Allow low-risk actions'. This change reflects usage patterns where over 70% of health data users have allowed their data to be utilized without being asked every time. Consequently, after connecting to health information, ChatGPT can use data in relevant contexts without asking for permission every time, except when attempting sensitive tasks such as emailing a workout plan to a running partner. Permissions can be changed at any time in the settings under the Plugins and Health menu.

Along with this, ChatGPT's memory feature has also been upgraded to more accurately maintain the user's latest context. It helps reduce outdated or contradictory stored memories while better understanding the user's preferences, goals, and ongoing tasks. ChatGPT automatically tracks details deemed important to update memories, and users can directly review the memories used for response personalization via the source or memory summary pages.

Image source: OpenAI