Get the weekly research letter
Companies change their terms quietly. We read every version and catch what actually changed. One email a week on the changes that matter and what they mean. No account.
When the Keep Activity setting is on, Google uses saved chats, uploaded files, videos, screens, photos, audio, Gemini Live recordings, feedback, website visit information, and location data to train generative AI models. Turning off Keep Activity prevents future chats from being used for AI model training unless the user submits feedback.
This analysis describes what Google Gemini's agreement states, permits, or reserves. It does not constitute a legal determination about enforceability. Regulatory applicability and practical outcomes may vary by jurisdiction, enforcement context, and individual circumstances. Read our methodology
This provision establishes that a broad range of data categories, including screen content, audio recordings, and uploaded files, are used for generative AI model training when Keep Activity is enabled, and that the default auto-delete period is 18 months. The notice discloses that even with Keep Activity off, feedback submissions trigger collection and use of the preceding 24 hours of chat context for model improvement purposes.
The updated privacy notice establishes that Google collects data from third-party services you connect to Gemini, including Model Context Protocol server tools, and that such connections are not monitored or secured by Google. The notice explicitly states that choosing to connect third-party apps may expose your data, passwords, devices, and accounts to unauthorized access. The revised terms also clarify that you can use temporary chats, which are not retained for AI improvement purposes with human reviewer assistance. You can manage connected app permissions through Gemini Spark settings.
View change record →The updated notice adds new disclosure sections explaining how data flows when using Gemini Spark (remote browser and computer access), how avatar creation collects and processes information, and clarifies that Google collects information about AI reasoning steps during task execution. The notice also refines language around subscription information to specify 'Google AI plan' rather than just generic 'paid subscription.' These changes do not establish new obligations but rather expand the transparency disclosures provided to users about existing and new features.
View change record →The updated notice now explicitly identifies Memory as a feature that operates on the basis of user consent, alongside Voice Match. The revised language removes prior geographic restrictions on personalization, meaning Gemini can now reference chat history to generate personalized insights for all users globally, not just those outside the EEA, Switzerland, and the UK. The removal of the statement 'Keep Activity must be on to use this feature' simplifies the operational requirement but does not establish a new obligation. You can learn how to turn the Memory feature on or off through the updated privacy notice.
View change record →Under this clause, users with Keep Activity enabled have their chats, uploads, audio, video, and screen content used to train Google's generative AI models, with an 18-month default auto-delete period that users can adjust to 3 months, 36 months, or indefinite. The agreement states that turning off Keep Activity stops future chats from being used for AI model training, but submitting feedback while Keep Activity is off re-enables collection of the preceding 24 hours of chat context for that feedback instance.
Cross-platform context
See how other platforms handle AI Model Training Using Keep Activity Data and similar clauses.
Compare across platforms →Monitoring
Google Gemini has changed this document before.
Receive same-day alerts, structured change summaries, and monitoring for up to 25 platforms.
"If the Keep Activity setting is on: Your chats and what you share with Gemini (like files, videos, screens, and photos) will be saved in your Activity. ... Google uses your activity to provide, develop, and improve its services (including training generative AI models), as well as to protect Google, its users, and the public with the help of human reviewers. ... If the Keep Activity setting is off: Your future chats won't appear in your Activity, and won't be used to train our AI models, unless you choose to send Google feedback.Excerpt from Google Gemini's Gemini Apps Privacy Notice
1. REGULATORY LANDSCAPE: This provision implicates GDPR legitimate interests and consent bases, as well as the EU AI Act's requirements regarding training data transparency for general-purpose AI models. The FTC and EU data protection authorities are relevant enforcement bodies. The notice asserts legitimate interests as the legal basis for AI model training using chat data, which requires a balancing test under GDPR that accounts for the sensitivity and volume of data categories involved. 2. GOVERNANCE EXPOSURE: High. The breadth of data categories used for AI model training, including screen content, audio recordings, and Connected Apps data, and the default opt-in nature of the Keep Activity setting, create significant compliance exposure under GDPR legitimate interests assessments and emerging AI governance frameworks. 3. JURISDICTION FLAGS: EU member states and the UK face heightened exposure due to explicit GDPR applicability. The EU AI Act may impose additional transparency and data governance obligations depending on how Gemini's models are classified. California users may have CCPA rights relevant to this use of personal data for AI training purposes. 4. CONTRACT AND VENDOR IMPLICATIONS: Organizations deploying Gemini where employee or customer data may enter the system should assess whether AI model training use of that data is disclosed in their own privacy notices and whether it is consistent with their data processing agreements. The notice does not indicate that enterprise or Workspace accounts are subject to the same training data terms, but the document states that work or school accounts may be subject to different terms. 5. COMPLIANCE CONSIDERATIONS: Legal teams should assess whether the legitimate interests basis asserted for AI model training satisfies GDPR balancing test requirements given the volume and sensitivity of data involved. Organizations should review whether Keep Activity settings are configured consistently with their data governance obligations and whether employee use of Gemini Apps is addressed in internal data handling policies.
Full institutional analysis
Regulatory citations, enforcement risk, and due diligence action items.
Monitor: same-day alerts on the platforms you choose. Analyst: full institutional analysis.
Compliance Governance Intelligence
Need to monitor specific governance provisions?
Compliance includes provision-level monitoring, governance timelines, regulatory mapping, and audit-ready analysis.
Built from archived source documents, structured governance mappings, and historical version tracking.
This provision establishes that a broad range of data categories, including screen content, audio recordings, and uploaded files, are used for generative AI model training when Keep Activity is enabled, and that the default auto-delete period is 18 months. The notice discloses that even with Keep Activity off, feedback submissions trigger collection and use of the preceding 24 hours of …
Under this clause, users with Keep Activity enabled have their chats, uploads, audio, video, and screen content used to train Google's generative AI models, with an 18-month default auto-delete period that users can adjust to 3 months, 36 months, or indefinite. The agreement states that turning off Keep Activity stops future chats from being used for AI model training, but …
No. ConductAtlas is an independent monitoring service. We are not affiliated with, endorsed by, or sponsored by Google Gemini.