Google will extend the temporary retention period for Gemini chats and conversations generated when chat history is not saved, to about two months. The company stated that this move is aimed at enhancing security and protection levels as the capabilities of the AI system are strengthened.
This adjustment will be globally rolled out at Gemini in the coming weeks. According to an archived version from November 12th, Google previously mentioned a retention period of 72 hours for Gemini Apps Privacy Hub, so this adjustment means that the retention period has been significantly extended.
In contrast, OpenAI indicates that temporary chats in ChatGPT may be retained for up to 30 days; Anthropic defaults to retaining invisible chats in Claude for 30 days as well. However, for legal or security reasons, or to investigate policy violations, relevant companies may retain some data for an extended period of time.
Google did not answer questions regarding the reasons for this adjustment, or whether it would affect the way conversations are accessed or reviewed. Instead, they cited a blog post. The post stated that extending the chat retention period is to respond to users, handle feedback, and protect their services and users. The updated Gemini Apps Privacy Hub also lists potential system failures as one of the reasons for retaining chat records, and indicates that temporary conversations will not be used to train the AI model.
As longer retention periods are introduced, Google is also launching separate chat history and training settings for Gemini and AI. Previously, these two functions were tied under a single setting called Keep Activity. In the future, users will be able to retain their conversations in the chat history without allowing this content to be used to train Google's generative AI models.
Google stated that for users who previously had " Keep Activity " disabled, these two new settings will by default remain disabled. For users who previously had " Keep Activity " enabled, their chat history will still be retained. However, they can decide individually whether to allow Google to use their conversations to train the " AI " model.











