Google Gemini learns from chats; new privacy controls offered
As Google’s Gemini chatbot becomes increasingly integrated into the company’s ecosystem, its interaction with user data is undergoing significant shifts. Google recently announced several key updates to how its artificial intelligence models adapt to individual users, notably introducing the ability for Gemini to retain more details from past conversations to generate improved responses. For those concerned about privacy, a new temporary chat option offers a way to interact with Gemini without influencing its future understanding of your preferences.
This latest customization effort, dubbed “Personal Context,” marks a new approach after Google previously experimented with a “personalization” feature that drew on users’ search history – a concept that apparently didn’t resonate widely. When Personal Context is enabled, Gemini will proactively recall specifics from previous interactions, tailoring its replies without explicit prompting. Google asserts that this will lead to more pertinent responses, particularly for recommendation-based queries. This feature operates distinctly from the “saved instructions” option, where users provide explicit directives for Gemini to follow in its outputs. While such personalization could make Gemini feel more engaging, it also carries a risk: AI chatbots that become overly familiar with users have, in the past, been observed to reinforce misconceptions or even foster delusional thinking.
Initially, Personal Context will be available exclusively with the Gemini 2.5 Pro model for users over the age of 18, and it will not be rolled out in the European Union, the United Kingdom, or Switzerland. Google plans to expand this functionality to additional regions and eventually include support for the more efficient Gemini 2.5 Flash model. Users will retain full control, with the ability to toggle Personal Context on or off at any time from the main settings page.
Alongside Personal Context, Google is rolling out “Temporary Chats,” a feature akin to an Incognito Mode for Gemini. This provides a crucial option for users who might hesitate to engage in certain conversations with the chatbot due to privacy concerns. Temporary Chats, which are also beginning their rollout and will become available to all users in the coming weeks, can be accessed via a dedicated button next to the “New chat” option within the Gemini application. Any interaction conducted in a temporary chat will not be used in Personal Context, even if that setting is enabled. While Google describes these as “one-off” chats, they are not immediately ephemeral; they will be retained on Google’s servers for 72 hours, allowing users a brief window to refer back to or expand upon the conversation if desired.
Beyond these new personalization features, Google has confirmed a more profound change in how it utilizes content uploaded to Gemini. Starting September 2, a sample of user chats and uploaded data—including files—will be used to train Google’s AI models. Google states this is to “improve Google services for everyone.” Users who wish to prevent their data from being incorporated into Google’s AI training models will need to actively opt out. In the coming weeks, Google will update its account-level privacy settings, renaming “Gemini Apps Activity” to “Keep Activity.” Disabling this setting, or consistently using Temporary Chats, will prevent your data from being used in Google’s model development. It is advisable for all Gemini users to review this crucial setting before September to ensure their data privacy preferences are respected.