Updated
Updated · Ars Technica · Nov 17
Google Separates Gemini Chat History From AI Training, Retaining Data for Up to 3 Years
Updated
Updated · Ars Technica · Nov 17

Google Separates Gemini Chat History From AI Training, Retaining Data for Up to 3 Years

3 articles · Updated · Ars Technica · Nov 17

Summary

  • Google is rolling out separate Gemini toggles for chat history and model training, ending the previous requirement to disable history to stop chats and attachments from being used for AI training.
  • The new setup lets users keep basic chat continuity while opting out of future training, but training remains enabled by default and must be turned off in Gemini settings or Google account activity controls.
  • Future interactions only are covered by the change; earlier prompts, responses, uploaded files, tool calls, location data and connected-app data already in Gemini activity can still be kept for up to three years.
  • Google says it tries to filter personal information from Gemini activity history, but the company does not guarantee that personal data is fully excluded from training-related records.

Insights

Even with Google's new privacy toggle, what happens to the personal data Gemini has already permanently absorbed from your past chats?
If human reviewers can still access your Gemini conversations for up to three years, is your private data ever truly safe?