Google Extends Gemini Chat History Retention to Two Months, Raising Privacy Concerns

Google extends Gemini's default chat history retention from 72 hours to two months, even if history is turned off.
Google announced it will significantly extend the default retention period for Gemini chat records from 72 hours to two months, and this rule applies to users who turn off chat history. The update will gradually roll out globally over the coming weeks. Google says the move aims to improve service quality and model training effectiveness, but for privacy-conscious users it means "turning off history" no longer equals conversations not being retained. The article explains the impact of this change on incognito use expectations and recommends that users view and manage chat data through Google account privacy settings and manually delete unwanted conversations, while noting that manual deletion may not affect data already used for training. It also places this in the broader trend of tightening data retention policies across the AI industry and reminds users to proactively review privacy settings.
Google recently announced that it will extend the default retention period for Gemini chat records from 72 hours to two months, even if you turn off chat history. This change will roll out globally over the coming weeks.

Retention Period Greatly Extended, User Privacy Faces New Considerations
According to Google's explanation, this adjustment significantly extends the original 72-hour chat retention period to two months. This means that even if users choose to turn off chat history, Gemini will still store conversation content for up to 60 days. Google says this change is intended to improve service quality and model training effectiveness.
For privacy-conscious users, this is undoubtedly a change worth paying attention to. While Google emphasizes that retained data is used to improve services, a longer retention period also increases the risk of data leaks or improper access.
The Chat History feature is a common user-controllable setting in AI assistants such as Gemini. When enabled, users' conversations are saved to their account, making them easy to trace across sessions and personalize; when disabled, it traditionally meant conversations would not be stored long-term. Google's previous commitment for the off state was to retain data for only 72 hours for security and abuse detection, then delete it. This adjustment extends the retention period in the off state to 60 days, essentially changing the user expectation that "off means no trace," making the switch more of an option for "whether to use for personalized display" rather than a switch for whether data is stored. Understanding this semantic change is key to judging the privacy impact.
Use of data for model training and the retention period are two different levels. The retention period refers to how long conversation content is stored on Google's servers; whether it is used to train models depends on the terms of service and the "improve services" option in user settings. Even if chat records are deleted, data that has already entered the training process is usually difficult to retroactively remove, because the model parameters have already absorbed this information. Therefore, extending the retention period expands the data pool that can be used for training or manual review, but manually deleting conversations does not necessarily equal "never used for training." Users concerned about this need to check both the storage period and the training authorization settings.
Global Rollout, Users Need to Understand the Impact
This update will gradually roll out to users worldwide over the coming weeks. Google did not provide a specific phased timeline but said it will ensure a smooth transition. Users should note that even if history is turned off, chat content will still be retained for two months, which may affect users who rely on "incognito" mode for sensitive conversations.
How to Manage Your Gemini Chat Data
Users can view and manage Gemini chat records through the privacy settings of their Google account. Google recommends that users regularly check their privacy settings and manually delete conversations they do not want to keep as needed. However, it is currently unclear whether manual deletion will affect data already used for training.
Industry Trend: AI Chat Retention Periods Are Generally Being Extended
This is not an isolated case. As AI assistant capabilities become increasingly powerful, major vendors are adjusting data retention policies to balance service improvement and user privacy. Google's move may prompt other vendors to follow suit, but it may also push regulators to strengthen scrutiny of AI data practices.
For ordinary users, understanding these policy changes and proactively managing their own data is crucial. While enjoying the convenience of AI, it is wise to stay attentive to privacy settings.
The meaning of "incognito mode" in AI chat scenarios is not the same as browser incognito. Browser incognito mainly avoids local record retention, while AI assistant conversations still need to be transmitted to the cloud for inference, and the server inevitably comes into brief contact with the content. The industry usually uses "short-term retention after turning off history" as a compromise for security auditing and troubleshooting. When the retention period is extended from hours to dozens of days, the privacy boundary of incognito conversations narrows accordingly. If users truly have sensitive content needs, a more reliable approach is to avoid entering personally identifiable information in conversations rather than relying solely on the history toggle.
Summary
Google's extension of Gemini chat record retention from 72 hours to two months is another signal of tightening data policies for global AI services. Users should review their privacy settings, understand how data is used, and take appropriate measures to protect personal information.
Related articles

Snapchat Opens AI Assistant Lens Building: Supports Claude Code and Cursor
Snapchat lets Lens creators describe AR filters in natural language, with AI assistants like Claude Code and Cursor handling the full build process, greatly lowering the AR creation barrier.

Lenovo Legion Go Beam Leak: Could Be the Most Affordable Handheld
Lenovo Legion Go Beam handheld leaks, possibly positioned more affordably. This article analyzes Lenovo's current handheld pricing, the potential positioning of the Legion Go Beam, and why cheap handhelds are a key market gap.

Altman on Garfield Playing Him: Life Is Too Weird, This Isn't Even Top 10
OpenAI CEO Sam Altman responds to Andrew Garfield playing him in a film, saying it doesn't rank in his top 10 weirdest life moments. We analyze the symbolism behind the casting and the trend of tech leaders being adapted for entertainment.