Google is extending how long it keeps temporary Gemini conversations and chats made with history turned off, from 72 hours to about two months. The change is part of a broader update that separates the controls for saving conversations in the app and allowing Google to use them for training its generative AI models. The privacy notice describes the new retention period and the purposes for which the company says it uses the data.

The changes are rolling out globally in the Gemini app over the coming weeks, TechCrunch reported. Some users may therefore see the old settings while others already see separate controls. The updated policy gives users a way to preserve a history they can return to while declining one use of their content. For people who turn history off entirely, it lengthens the time that Google retains conversations on its systems.

Chat history and training become separate choices

Under the earlier setup, users who wanted Google to stop using ordinary chat content for training had to turn off the activity setting. That also took away the ability to return to past chats through the usual history view, Ars Technica reported. Under the new setup, people can keep their history and turn off the separate chat-content setting for model training.

For someone who uses the app to work on a recurring project, that difference means being able to return to a thread without agreeing to let Google use future content from that thread for training. Google’s guidance says chat content includes prompts and attachments. Other uses described in the notice, including providing the service and protecting users, can continue even when users turn the training control off.

The setting addresses future interactions. A user who changes it should not expect the switch to remove material already used for model training. Ars Technica reported that content already reviewed for training can be subject to a separate retention period. Deleting a conversation from visible history and changing how new conversations are handled are separate actions.

What the longer period covers

Google initially described temporary chats as a way to ask questions without adding the discussion to recent chats or Gemini activity, and said it would retain those conversations for up to 72 hours. The revised privacy guidance says the company will retain those chats and chats made with history off for around two months. The Register corroborated the earlier period by comparing the revised documentation with an archived version.

Google says it uses the last 24 hours of a chat as context for responding. The roughly two-month period governs how long the data can remain associated with an account, including to assist with safety and security and to make it available in the event of a system failure. Those two time frames describe different uses of the same information. The updated notice does not say that Gemini will use two months of conversation as context in an ordinary reply.

The company says temporary chats are excluded from model training. If history is off and the user does not submit feedback, Google also says it does not use future chats to train its generative models. Its data policy still allows human review for protection and safety. The policy describes human review as a process that can serve purposes beyond training, including responding to potential service and policy violations.

Sending feedback carries separate terms

The policy treats voluntary feedback differently. If users send feedback with history off, Google says it can collect the feedback, relevant context from the previous 24 hours and material in those chats such as files or information from connected apps. The feedback section says trained teams may review the material and that reviewed feedback and associated data can be kept for up to three years, disconnected from the account. That period applies to reviewed feedback and related data, while the regular history-off rule addresses ordinary retention.

History can include more than the text a person types. The privacy hub lists responses, attachments, audio and some content from services connected to Gemini, depending on how the app is used. Those categories matter when a user decides whether a given conversation should be retained or used for training. The availability of separate controls lets users decide what stays accessible in the account without making that same decision for every use of the content.

What users will see as settings roll out

The updated interface gives users two controls where the older one combined decisions. Ars Technica reported that the separate training option appears under activity settings as the change reaches accounts. People who use temporary conversations will see less change in their visible history than in the policy governing how long those exchanges stay on Google’s systems. The company’s revised documentation sets the new terms, while account-by-account rollout determines when users can make the two choices separately.