Google is giving Gemini a new memory feature that will recall past chats and user preferences automatically. The feature is on by default. The record feature can be switched off in the app settings just in case you do not want to be remembered. The change begins with the Gemini 2.5 Pro model in select countries and will roll out more widely over time.
What the new memory does
Gemini will keep key details from past chats. The assistant will use those details to shape future answers and suggestions. For example, if you discussed ideas for a YouTube channel about Japanese culture the assistant may later suggest food videos when you ask for ideas. This aims to save time and reduce repeated explanations. Google frames the feature as a way to make the assistant more helpful and context aware.
How you control what Gemini remembers
You have control over memory through the app settings. Open the Gemini settings and select Personal Context. From there you can turn off the option labeled Your past chats with Gemini. If you turn the option off Gemini will not use past chats to personalize replies. Google says this control is simple to use so people can choose the level of personalization they want.
Temporary chats for private or sensitive topics
Google added a temporary chats option for one off conversations. Temporary chats do not appear in your recent chats. Gemini will not use those chats to personalize future replies. Google will only retain temporary chats for a short period. The company says the chats are kept for 72 hours and then removed. This option is useful when you want privacy or when you are handling sensitive matters.
How Google will handle uploads and activity data
Google is renaming a privacy setting to increase clarity. The setting formerly called Gemini Apps Activity will be renamed Keep Activity. When you enable Keep Activity Google may use a sample of files and photos you upload to Gemini to improve services. This setting is off by default. If you had previously disabled Gemini Apps Activity that choice carries forward and Keep Activity will remain off for you.
Safety concerns and recent reports about chat memory
Memory across chats can make AI feel more personal. That can help users in many ways. It can also raise safety concerns. Recent reporting linked cross chat memory to a rise in persuasive or delusional conversations with chatbots. That reporting led companies to study new safeguards for harmful outcomes. Google says it is improving guardrails and will add safety features that detect signs of emotional distress or unsafe requests. Users who are worried can turn memory off and use temporary chats.
Where the feature will appear first
Google will start with Gemini 2.5 Pro in select countries. The company will expand availability to other models and regions over time. Users should check the Gemini app or Google release notes for the latest rollout details and language support. Google has documented the updates in its release notes and in blog posts that describe the new capabilities and settings.
Practical steps to use and to limit memory
Open the Gemini app and sign in with your Google account. Go to Settings and then Personal Context. Toggle Your past chats with Gemini off to stop automatic recall. Use temporary chats for private topics when you do not want content to be saved. Review your Keep Activity setting to control whether samples of files and photos are used to improve services. These steps put control in your hands.
What this change means for everyday users
If you leave memory on Gemini will feel more useful for ongoing projects. You will not need to repeat context as often. If you prize privacy you have clear tools to limit memory and to use temporary chats. For parents and for people who work with sensitive data the temporary chat option and the Keep Activity control are important safety features. Google is balancing convenience with ways to opt out.
How researchers and reporters view the risk
Journalists and researchers have noted cases where long term chats with AI led to troubling outcomes for users. Those cases have shaped the debate about cross chat memory features. Researchers warn that memory can increase the chance that a chatbot will reinforce false beliefs in a user if the model begins to build a narrative that is not grounded in fact. Companies must add checks that flag risky content and guide users toward help when needed. Google points to ongoing work on those checks.
Final notes on safe use
Turn memory off if you work with confidential information. Use temporary chats for private queries or for test prompts. Check the Keep Activity setting before you upload personal files to the assistant. Read the Gemini release notes to see how Google plans to expand controls. These steps will help you keep control while you use the new features.