· via TechCrunch
Google to retain history-disabled Gemini chats for two months instead of 72 hours
Google will keep Gemini conversations made with chat history off for about two months instead of 72 hours, citing safety, while splitting chat history and AI training into separate settings.

Google will now hold on to Gemini conversations started with chat history switched off for roughly two months, replacing a retention window of just 72 hours. According to TechCrunch, the change is rolling out globally in the Gemini app over the coming weeks.
The company frames the longer retention as a safety measure that has become necessary as its AI systems grow more capable. For users, it sharply redefines what disabling history means in practice: those chats stay invisible in the app, but they now persist on Google's servers about twenty times longer than before.
From three days to two months
An archived copy of Google's Gemini Apps Privacy Hub dated November 12 still listed the old 72-hour period, TechCrunch reports. The updated policy extends retention to around two months, and it applies both to temporary Gemini chats and to conversations conducted with history turned off.
What Google says, and what it won't
TechCrunch asked Google what prompted the change and whether it affects how conversations are accessed or reviewed internally. The company declined to answer, instead pointing to a blog post that says chats are kept longer so Google can reply to users, act on feedback and defend its platform and its users. The revised privacy hub also cites possible system failures as a justification for retention, and states that temporary conversations do not feed into AI model training.
How that compares with rivals
The new window is the longest of the disclosed defaults among the major chatbot providers. OpenAI says temporary ChatGPT conversations may be kept for up to 30 days, while Anthropic retains incognito Claude chats for 30 days by default. All of these companies note that some data can be held even longer for legal or safety reasons, or while investigating possible policy violations.
History and training are being decoupled
The retention change arrives alongside a rework of Gemini's privacy controls. Until now, a single "Keep Activity" setting governed two distinct things: whether chats were saved to your history, and whether Google could use them to train its generative AI models.
Those functions are now being separated. Users will be able to keep conversations in their chat history while excluding them from model training, a combination that was impossible under the old unified toggle.
According to TechCrunch, both new settings will be switched off by default for anyone who had previously disabled Keep Activity. People who had it enabled will keep their chat history on, and can then separately decide whether Google may use their conversations to train its models.
Why it matters
The core privacy expectation behind a "history off" toggle is that a conversation is short-lived. Google's move from three days to roughly two months breaks that assumption: the chat disappears from your view but remains in the company's hands for weeks, potentially used to respond to you, handle feedback or investigate abuse.
Because Google would not clarify whether the change affects human review or internal access to those conversations, users cannot fully evaluate the tradeoff they are being asked to accept. The only firm reassurance is that these temporary chats are excluded from AI training.
At the same time, the split settings are a genuine improvement in another respect, giving users a middle ground between full activity tracking and total ephemerality. The move also signals a broader industry drift: as chatbots take on more capable and consequential tasks, vendors from OpenAI to Anthropic to Google are quietly extending how long even supposedly throwaway conversations live on their servers.
- #gemini
- #privacy
- #ai
- #chatbots