Google recently announced a major privacy update that separates Gemini chat history from AI model training. Specifically, the new Gemini chat history settings resolve a widespread complaint among users who wanted to save conversations without authorizing model training. Now, individuals can preserve their conversation logs while preventing Google from using chat text and attachments to train future models. Google confirmed that these granular privacy controls will roll out to global users over the coming weeks.
Independent Toggles for Chat History and Model Training
Previously, a single “Keep Activity” setting governed both conversation storage and AI training simultaneously. Consequently, users who wanted to review past chats had to allow Google to train models on their prompts. Under the updated architecture, a dedicated History toggle manages saving and organizing conversation records. Meanwhile, a separate privacy switch lets users disable data authorization for AI model training.
As detailed in the official Google announcement on updated Gemini privacy controls, users can now mix these options freely. For individuals who previously disabled Keep Activity, Google sets both new toggles to off by default. However, those users can easily re-enable the History option without opting back into model training.
Data Retention Rules When History Is Disabled
Importantly, Google reminds users that disabling chat history does not trigger immediate deletion from backend servers. When History remains off, new conversations disappear from the visible chat interface. Nevertheless, Google retains those prompts within the user’s account for roughly two months to deliver responses, process feedback, and protect AI services.
Human Review and Enterprise Workspace Controls
Furthermore, opting out of model training does not stop Gemini from processing prompts to generate answers or enforce safety filters. If a user submits explicit feedback, trained reviewers may still inspect the relevant conversation to improve service quality. According to Google, human-reviewed conversations and associated metadata remain stored for up to three years after decoupling from user accounts.
Finally, Google Workspace administrators continue to manage history permissions centrally for work and school accounts. Users should also note that turning off chat history may limit the availability of certain connected applications. Ultimately, decoupling conversation logs from model training grants everyday users far stronger control over their personal AI data.
Support Our Threat Intelligence
Find our tech and OS security coverage helpful? Support our work today and unlock a 100% ad-free reading experience!