AI NewsInfrastructureAnnouncement
Google now keeps Gemini chats for two months with history off
Google updated the Gemini privacy page. Temporary chats and history-off chats now stay for about two months. The old number was 72 hours.

Why it mattersA team that sends prompts or customer data to Gemini should now assume Google keeps it for about 60 days.
Turning chat history off used to mean Google kept the chat for three days. It now means Google keeps it for about two months. Google updated the Gemini privacy page on November 17 to say so.
The page now names two kinds of chat that stay for around two months. The first kind is a Temporary Chat. The second is any chat sent while History is off. Both are kept with the user's account for that window.
TechCrunch's Jagmeet Singh noticed the change by comparing the live page to a November 12 copy in the Wayback Machine. The old wording said 72 hours. The new wording says about two months. Google would not answer TechCrunch's question about what prompted the change. Google pointed to the same page. The page says the longer window is to respond to users, process feedback and protect the service.
What is in scope
The two-month window covers chats a user told Gemini not to keep. Google says these chats stay out of model training.
Chats from users who leave History on get the retention they chose: 18 months by default, with 3 months, 36 months or indefinite as options. Those are used for training and are reviewed by humans. The page also says reviewed chats, once disconnected from an account, can stay for up to three years. That window applies even after a user deletes their history.
Google separated the training setting from the history setting at the same time. A user who had History off now has training off as well, by default. Turning off training does not stop a chat from being reviewed for abuse or for service quality.
What a team should do with it
The group that pays for this policy is the team running an app that calls Gemini. Any prompt sent through Gemini is stored with Google for about two months. That is true even on the privacy setting that looks strictest from the user's side.
A team that sends regulated data through Gemini needs to ask one question on the next review. Is a sixty-day window inside the retention promise the team made to its own customers? Health, financial and legal data are the obvious cases.
A team that lets users reach a Gemini-backed feature in Temporary Chat mode needs to ask a second question. Does the help text accurately tell the user what Google actually keeps?
The field has three published numbers now. OpenAI says temporary ChatGPT chats are kept up to 30 days. Anthropic says incognito Claude chats are kept similarly. Google's new figure is roughly twice either. TechCrunch flags this as the main shift the change represents.
The old 72-hour wording was a policy choice, not a technical limit. A user can delete a chat sooner and Google will honour that. The visible change is what happens by default. That default is now two months.
Source
Gemini Apps Privacy Hub, Google. TechCrunch's report sets the before-and-after wording through a Wayback Machine comparison.
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.
