Google is separating Gemini chat history from AI training controls, allowing users to save conversations without using their data to train AI models. Temporary chats will now be retained for around two months instead of 72 hours.
Google is introducing new privacy controls for its Gemini artificial intelligence (AI) app, giving users greater flexibility over how their conversations are stored and whether their personal data is used to train AI models. The update separates chat history from AI training preferences, allowing users to retain access to previous conversations without automatically contributing their interactions to Google’s AI development.
The change addresses a growing concern among AI users who want the convenience of saved conversations but are uncomfortable with their personal prompts, uploaded files and other interactions being used to improve AI systems.
Until now, Google’s Gemini app had a single setting called ‘Keep Activity’, which controlled both the storage of conversations and the use of that information for generative AI model training. Users who wanted to maintain their chat history also had to allow Google to use their interactions for improving its AI models, depending on their settings.
With the latest update, Google is splitting these functions into two independent controls. The first, called ‘History’, determines whether conversations are saved and how long users can access them. The second allows users to decide separately whether their chat content can be used to train Google’s generative AI models.
This means Gemini users will no longer have to choose between keeping their conversations and protecting their data from AI training.
Two separate privacy controls
The revised settings are expected to give users a clearer understanding of how their information is handled when interacting with Gemini. Users can keep their conversation history enabled while switching off the option that allows Google to use their content for generative AI training.
The training preference covers more than just text prompts. According to Google’s updated privacy documentation, it can include audio transcripts, photographs, documents, videos, screen content, Gemini’s responses, tasks and tool interactions, as well as certain information from connected applications.
When the AI training option is disabled, future conversations will not be used to train Google’s generative AI models unless users voluntarily submit feedback. Google, however, says it may continue reviewing conversations and using relevant information to improve its services, maintain security and address potential problems.
The distinction is important because disabling AI training does not mean that every interaction becomes inaccessible to Google or is immediately deleted from its systems.
Gemini chats will now be retained for two months
Alongside the new privacy controls, Google is also changing how long it retains conversations when users switch off their chat history.
Under the earlier policy, temporary chats and conversations created with Keep Activity disabled were retained for approximately 72 hours. The revised policy extends this period to around two months, marking a substantial increase in the duration for which such conversations remain stored.
Google says the longer retention period is necessary to support safety and security measures, process user feedback and protect the company, its users and the wider public. The company has also referred to the need to retain conversations in case of potential system failures.
Importantly, chats created in temporary mode are not used to train Google’s AI models. Similarly, when History is switched off, future conversations are not used for AI training unless users submit feedback.
However, the two-month retention period means that switching off chat history does not amount to an immediate deletion of conversations from Google’s systems.
Google’s updated privacy documentation also explains that the last 24 hours of a conversation may be used as context to respond to users. This is separate from the longer retention period for safety and security purposes.
What happens when users submit feedback?
There is another important privacy consideration for Gemini users.
If users submit feedback while History is disabled, Google may collect the feedback along with relevant conversation details, including the last 24 hours of chats, attachments and information from connected applications.
Such material can be reviewed by specially trained human reviewers to investigate reported problems, identify potentially unsafe responses and improve Gemini’s performance.
Google says conversations and related information reviewed through this feedback process may be retained for up to three years, even if users subsequently delete their activity or change their privacy settings.
The company also states that when users enable AI training, their conversations and associated information may be shared with service providers to help improve generative AI models. This information is separated from their Google accounts before being sent for review.
Consequently, users should understand that switching off the training preference prevents future conversations from being used for model training, but does not necessarily reverse the treatment of information already reviewed or processed under earlier settings.
When will the new Gemini settings arrive?
Google announced the changes on November 16 and said the revised privacy controls would begin rolling out globally through the Gemini app over the coming weeks. Availability may therefore vary across accounts and regions during the rollout.
Users who had previously switched off Keep Activity will have both new settings disabled by default, according to Google. Those who had the earlier setting enabled will retain their chat history, but can independently decide whether their conversations can be used for AI training.
For users who keep History enabled, Gemini conversations are subject to the existing automatic deletion schedule, which defaults to 18 months. Google allows users to change this period to three months, 36 months or indefinite storage. Conversations can also be deleted manually.
The update comes as technology companies face increasing scrutiny over how they collect, retain and use personal information to develop artificial intelligence products.
For everyday Gemini users, the change offers a more practical approach to privacy. People who rely on the chatbot for writing, research, planning and other routine tasks can continue accessing previous conversations without necessarily allowing those interactions to become part of Google’s AI training process.
However, the extended retention period is a reminder that switching off chat history does not mean conversations disappear immediately. Users will need to examine both settings and understand the difference between saved history, temporary data retention and AI model training before deciding how much information they want to share with the chatbot.